You may remember how that ended. Nasdaq moved to delist the company the following year, and in 2021 the SEC charged three people with insider trading on the news of the pivot. The iced tea, as far as I know, was fine.

I think about Long Blockchain more than most, mostly because of the names that land on my desk. My firm clears a lot of names for entreprneurs, brands, and others, and over the past year or two an extraordinary share of them have had the letters AI in them somewhere, bolted on as a suffix, folded into a portmanteau or sitting at the end of the domain. When I mention that the name may be hard to protect, I usually get the same answer, delivered with a trace of impatience. “The investors need to know what we do.”

They do. That is exactly the problem, and it turns out to be a more interesting problem than it first appears, because the market and the law are measuring the same property of these names and pricing it in opposite directions.

* * *

Three finance professors noticed the market’s half of this a quarter century ago. In 2001, Michael Cooper, Orlin Dimitrov and P. Raghavendra Rau published a paper in The Journal of Finance titled “A Rose.com by Any Other Name,” which remains the best title any finance paper has ever had. They looked at 95 public companies that added “.com” to their names in 1998 and 1999 and found abnormal returns of roughly 74 percent in the ten days around the announcement. The detail that should have embarrassed everyone was that the bump seemed to have little to do with how much business a company actually did on the internet. Investors were buying the suffix.

What they were paying for, I think, was membership. A .com name in 1999 told the market that a company belonged to the category everyone wanted exposure to, and it said so in the most legible way available, by using the same word as everyone else. The value came from the sharing. Pets.com and Garden.com were announcing that they were part of the same thing, and the market loved them for it.

Trademark law is built on the opposite instinct. A mark exists to tell you who made something, and the law protects it only to the extent that it does that job. The framework courts still use comes from a 1976 opinion by Judge Henry Friendly in a dispute over the word “Safari” on Abercrombie & Fitch clothing, and its logic is simple enough to fit on an index card: the more a word tells you about the product, the less it can tell you about the source. “Kodak” tells you nothing about cameras, which is why it was so easy to protect. “AI” tells you a great deal about software. A word that an entire industry is using to describe itself is a word no single member of that industry gets to own.

The easiest illustration is also the funniest. OpenAI did more than any company on earth to turn three letters into a household term, and when it applied to register GPT, the Patent and Trademark Office said no. Its final refusal, in February 2024, called the term merely descriptive, since a generative pre-trained transformer is a type of technology that other companies build too. OpenAI’s success at popularizing the phrase only strengthened the case against it. The better the company did at making GPT mean something, the less GPT could mean OpenAI.

Most startups will never have OpenAI’s problem, and theirs is slightly different. An examiner will usually let a name like FooAI register, but the letters carry no weight in it, so whatever protection the mark has collapses onto the other word. With thousands of companies now attaching the same two letters to their own other words, a dispute between any two of them comes down to whether the remaining words are too close, decided against a background crowded with lookalikes. Courts have said fairly bluntly that a crowded field narrows everyone’s rights. It is entirely possible to spend a seed round building a brand whose legally meaningful part turns out to be a single ordinary English noun.

* * *

There is one party for whom all of this is unambiguously good business, and it is a Caribbean island of about fifteen thousand people. The .ai domain is the country code for Anguilla, assigned in the 1990s when the letters stood for nothing in particular, and the island has spent the last few years watching Silicon Valley line up to rent its initials. According to local reporting, Anguilla took in roughly $85 million from .ai registrations in 2025, more than double the year before, and its government expected the domain to produce nearly half of its total revenue. Registrations passed a million in January 2026.

Anguilla also illustrates something founders get wrong constantly. Trademark law treats a top-level domain as having about as much brand significance as a zip code. Owning acme.ai gives you a place to put a website. Plenty of founders assume it settled the question of the name, and some of them find out otherwise from a letter sent by a company with an earlier trademark filing and a perfectly vanilla .com.

* * *

I should be fair to the other side, because the history here does not run in one direction. Some of the most valuable names in business began life as fashionable category words. Microsoft is a contraction of “microcomputer software,” which in 1975 was about as descriptive and as trendy as a name could be. General Electric has carried the word “electric” for more than a century without anyone suggesting it change.

These names worked, and the reason they worked is the helpful part. The companies behind them won so completely that the shared word started pointing at them. Microsoft’s name became distinctive because its software ended up on most of the world’s computers, and trademark law simply recognized what the market had already done.

So the honest description of an AI name is that it is a bet on becoming the category, or at least on becoming so dominant within it that the common word bends toward you. In any boom, one or two companies win that bet. Everyone else ends up with a name that ages alongside the trend. Radio Corporation of America spent decades doing far more interesting things than radio before formally becoming RCA in 1969. Dunkin’ Donuts dropped the doughnuts in 2019, having noticed that it mostly sold coffee. Many of the dot-com survivors quietly deleted the suffix after the crash, and there is some evidence that the market rewarded them for that too, though I would not put too much weight on any single study of investor mood in 2001.

There is a smaller thing that bothers me, and it is more temperamental than legal. When I started practicing, founders would tell me what problem they were trying to solve and get around to the business model eventually. Now the first two things I hear are the technology and the valuation, and the names have followed the pitch. A name that announces the tool tells me what a company was built with. It says nothing about what the company is for, and the companies I have watched last tended to know the second thing early.

Artificial intelligence will go the way of electricity and the internet, which is to say it will become so ordinary that announcing you use it will sound like announcing you have plumbing. When that happens, the names will date their companies as precisely as a year stamped on a coin. A few will have become the category and will wear the word comfortably, the way Microsoft does. The rest will hire an agency to help them forget it, and in Anguilla the renewal invoices will go out on schedule.

What made the scene funny, though, is the part the quiet luxury discourse tended to skip. The bag had no name printed on it in any way the audience could read from a couch, and millions of viewers recognized it anyway. The Burberry check, introduced as a coat lining in the 1920s, has been doing the work of a logo for decades. The show was mocking a brand for being too loud while proving, in the same breath, that a pattern can be as legible as a word.

That is the story of the last several years in luxury and consumer branding. The logo did not die. It migrated off the label and into the product itself, into colors and silhouettes and weaves and store layouts, and the most valuable brands in the world are now the ones you can identify from across a room without reading anything at all. The question for anyone building a brand, or advising one, is what the law actually protects once the name comes off.

A Short History of Being Recognized in the Dark

The desire to be known without words is older than modern trademark law. In 1915, the Coca-Cola Company, frustrated by imitators selling cola in bottles that looked like its own, backed a competition among glassmakers with a peculiar brief: design a bottle so distinctive that a person could identify it by touch alone, or recognize it from a shard of broken glass on the ground. The Root Glass Company of Terre Haute, Indiana won with the contour bottle. Legend has it that the designers, searching an encyclopedia for an image of the coca leaf or the kola nut, landed on an illustration of a cacao pod instead and borrowed its ribbed curves. The company protected the design first with a patent, filed in November 1915 with the Coca-Cola script deliberately left off to keep the project secret. Trademark protection came much later. The Patent and Trademark Office registered the bottle bearing the Coca-Cola name in 1960, and it took until 1977 for the shape alone, with no words on it at all, to earn a registration of its own.

Tiffany & Co. had been working on a similar problem for even longer. The robin’s egg blue it chose in 1845 for the cover of its Blue Book catalogue became so bound up with its identity that, according to a widely repeated newspaper account, Charles Lewis Tiffany refused to sell the empty boxes at any price, insisting they could come only with something purchased inside. Tiffany secured trademark protection for the color beginning in 1998, and in 2001 it worked with Pantone to create a private shade called 1837 Blue, after the year of the company’s founding. Few brands have turned a color into a piece of corporate autobiography so neatly.

The theme runs through the rest of the century. Owens-Corning dyed its fiberglass insulation pink in 1956, adopted the Pink Panther as its mascot in 1980, and won a landmark 1985 decision from the Federal Circuit holding that the color could serve as a trademark for insulation, a product that no customer would ever see once it disappeared behind drywall. UPS built an identity on brown. Hermès, which was founded as a harness maker in the same year as Tiffany, turned the Kelly and the Birkin into objects so recognizable that their outlines alone carry meaning, and in July 2025 the original Birkin, made for Jane Birkin herself, sold at Sotheby’s in Paris for about $10.1 million, a record for any handbag.

None of these companies abandoned their names. What they understood, well before anyone used the phrase “quiet luxury,” is that a name is something a customer reads, while a shape or a color is something a customer simply knows, and knowledge of the second kind tends to run deeper.

What the Law Protects When It Isn’t a Logo

American trademark law has always defined a trademark by what it does rather than what it looks like. Under the Lanham Act, a mark can be any word, name, symbol, or device that identifies the source of goods, and the Supreme Court has read “symbol” and “device” generously. The body of law that protects the overall look of a product or its packaging is called trade dress, and most of its modern shape was set by four Supreme Court decisions in a single decade.

The first, Two Pesos v. Taco Cabana in 1992, involved a pair of Mexican restaurant chains in Texas. Taco Cabana had developed a festive interior and exterior design, with bright colors, painted murals, and garage doors that opened its interior patio to the outside, and Two Pesos had copied it closely. The Court held that trade dress of this kind could be protected without proof that customers had come to associate it with a particular company, so long as the design was inherently distinctive, meaning unusual enough that it would naturally be understood as a signature. It was a significant expansion. Restaurants and retailers, and eventually Apple, which registered the layout of its stores as a trademark in 2013, came to see their physical environments as protectable brand assets rather than decoration.

Three years later, in Qualitex v. Jacobson Products, the Court took up a question that had divided lower courts for decades: whether a color by itself could function as a trademark. The case concerned a particular green-gold shade used on pads for dry-cleaning presses, which is about as unglamorous as trademark litigation gets. Justice Breyer, writing for a unanimous Court, held that color could qualify, but only once it had acquired what lawyers call secondary meaning, the point at which consumers see the color and think of a single source rather than of the product generally. That requirement is the reason Tiffany’s blue and Owens-Corning’s pink are protectable and most colors are not.

Then, in 2000, the Court pulled in the other direction. Wal-Mart Stores v. Samara Brothers concerned a line of children’s seersucker clothing decorated with appliqués of hearts and flowers, which Wal-Mart had sent to a manufacturer as photographs with instructions to produce copies. Justice Scalia drew a line that still governs the field. Product packaging, he reasoned, can be inherently distinctive, because consumers are accustomed to reading a package as a signal of who made the thing inside. The design of the product itself is different, because consumers generally assume a shape or pattern was chosen to make the product more useful or more attractive, not to tell them who made it. Product design, the Court held, can never be inherently distinctive and must always prove secondary meaning.

The final piece came in 2001 with TrafFix Devices v. Marketing Displays, which dealt with a dual-spring mechanism that kept road signs upright in the wind. The Court held that a functional feature, meaning one essential to the product’s use or purpose or one that affects its cost or quality, cannot be protected as trade dress at all, and that an expired utility patent covering the feature is strong evidence of functionality. Trademark rights can last forever, and the Court was unwilling to let companies use them to extend monopolies that patent law had deliberately limited.

Taken together, these cases produce a framework that anyone working with a non-logo brand asset has to navigate. Packaging and store environments can be protected relatively quickly if they are unusual enough. Product designs and colors have to earn their protection through years of use and consumer recognition. And anything that makes the product work better stays in the public domain for competitors to use, as does, under a doctrine called aesthetic functionality, any feature whose exclusive use would put competitors at a significant disadvantage unrelated to reputation.

That last idea sits at the center of the most famous non-logo trademark case of this century. Christian Louboutin began painting the soles of his shoes red in the early 1990s, reportedly after borrowing an assistant’s nail polish to brighten a prototype, and registered the red sole as a trademark in the United States in 2008. When Yves Saint Laurent released a line of monochrome shoes that were red all over, including the soles, Louboutin sued. The Second Circuit’s 2012 decision was a split outcome that both sides claimed as a win. The court held that the red sole was a valid mark, but only when it contrasted with the color of the upper, which meant YSL’s all-red shoe did not infringe. A color could belong to Louboutin as a signature, in other words, but not as a color.

The Other Toolkit

Trademark and trade dress are the tools most people associate with brand protection, but they are not the only ones, and for new products they are often not the right place to start, since secondary meaning takes time to develop and a knockoff can arrive in weeks.

Design patents protect the ornamental appearance of a functional item, for fifteen years from grant on applications filed since May 2015, and they require no proof that consumers recognize anything at all. They were at the heart of the long litigation between Apple and Samsung over the shape and interface of the iPhone, which reached the Supreme Court in 2016 on the question of how damages should be calculated when a design patent covers only one component of a larger product. They became more contested in 2024, when the Federal Circuit, sitting en banc in LKQ v. GM Global Technology, discarded a decades-old test that had made design patents unusually difficult to invalidate as obvious. The decision made them somewhat easier to challenge, but it did nothing to diminish their central advantage, which is that they protect a new design from the day they issue rather than years later. Coca-Cola understood this in 1915.

Copyright can play a role as well, though a narrower one than designers tend to hope. Copyright does not protect useful articles like clothing or furniture as such, but in Star Athletica v. Varsity Brands in 2017, the Supreme Court held that decorative features of a useful article, in that case the stripes and chevrons on cheerleading uniforms, can be protected if they can be imagined as artworks separate from the article itself. Courts outside the United States have been similarly cautious. In February 2025, Germany’s Federal Court of Justice ruled that Birkenstock’s classic sandal models did not qualify for copyright protection as works of applied art, concluding that their design did not reach the level of artistic creativity the law requires. It was a meaningful loss for a company whose sandals are arguably more recognizable than its name.

At the outer edges of trademark law sit what practitioners call non-traditional marks. NBC registered its three-note chime in 1950, the first purely audible mark in the United States, and Hasbro registered the scent of Play-Doh in 2018. Harley-Davidson spent six years trying to register the distinctive sound of its V-twin engine before withdrawing the application in 2000 in the face of opposition from competitors who pointed out, reasonably, that the sound was largely a product of the engine’s design. It is a useful reminder that functionality applies to the senses beyond sight, and that a brand cannot claim ownership of a quality its product simply happens to have.

Where It Gets Hard

The cases that succeed tend to become legend, but the more instructive ones are the brands that tried to protect something the market clearly associated with them and found the law less accommodating than they expected.

Bottega Veneta is the purest example of quiet luxury as a business model. In 1978 the company ran an advertising campaign built around the idea that its customers needed no initials other than their own, and its identity rests on the intrecciato weave, a pattern of interlaced leather strips that, by co-founder Renzo Zengiaro’s account, began as a workaround because the brand’s sewing machines could not handle thick hides. When Bottega applied to register the weave in 2007, the examiner refused on the ground that it was ornamental and aesthetically functional, since woven leather is a common design technique. Bottega won in 2013 only after narrowing its claim to strips of a particular width woven at a particular angle and assembling evidence of more than $22 million in American advertising and a weave that appeared on over 80 percent of its goods. Even then the victory had limits. An application covering footwear failed in 2019 after an opposition, and in Europe, the General Court upheld refusals to register handbag shapes incorporating the weave.

Others have fared worse, particularly in Europe, where a mark generally has to be shown to have acquired distinctiveness across the relevant territory. Nestlé held a European registration for the four-finger shape of the KitKat bar for more than a decade before the Court of Justice of the European Union ruled in 2018 that it had never proven recognition in every member state where that proof was required. Cadbury’s application to register a shade of purple failed before the English Court of Appeal in 2013 after Nestlé opposed it, not because consumers failed to associate the color with its chocolate but because the application claimed purple as either the whole color of the packaging or merely its “predominant” one, which the court found too vague to define what was actually being claimed. Adidas saw a registration for its three stripes invalidated by the EU General Court in 2019. In New York, it lost a 2023 jury trial against Thom Browne, whose four-bar stripe design the jury found did not infringe, and it has since lost both its appeal and a later attempt to reopen the judgment.

The lessons across these cases are consistent. Evidence of recognition must be substantial and gathered deliberately, since courts want surveys, unsolicited media coverage, and advertising that directs consumers to the feature itself. How a claim is described matters enormously, as Cadbury learned. And a signature that a company owns in one jurisdiction, or for one category of goods, may be up for grabs in another.

What This Means for Brands Building Now

For a founder or a creative director, the practical implications are less about litigation than about the habits a company forms long before anyone files a lawsuit.

The first is to think about protection in sequence rather than all at once. A new product with a distinctive design should usually be considered for a design patent before or shortly after its public release, since the United States allows a one-year grace period after disclosure and grace periods elsewhere vary, with some jurisdictions considerably stricter. Trade dress protection for the same design will come later, if it comes at all, and only after the market has had time to learn it. Relying on trademark law alone for a new silhouette is, in most cases, relying on rights that do not yet exist.

The second is to build the evidentiary record while building the brand. The companies that win trade dress cases are usually the ones whose marketing pointed at the feature they later needed to protect, which is why so many legacy brands ran “look for” campaigns directing consumers to a color, a stitch, or a shape. Consistency matters just as much, because a signature that changes every season is difficult to prove as a signature at all, and Bottega’s success depended in part on showing that the same weave ran across nearly everything it sold. Equally important is what a company avoids saying. Marketing copy that praises a feature for its comfort, durability, or efficiency can become evidence of functionality in the hands of a defendant, and more than one brand has found its own advertising quoted back to it in court.

None of this means the logo has lost its value. Far from it! Names remain the easiest things to register and enforce, and a word mark is still the backbone of nearly every brand portfolio. What has changed is that the most durable brands now carry their identity in more than one place, so that the name can go quiet without the brand going unrecognized. The scene in Succession worked because the audience knew exactly whose bag it was without reading a word, and the brands that will matter over the next generation are the ones that treat that kind of recognition as an asset to be documented, registered, and defended with the same care they once reserved for the name on the label.

On Thursday, a federal appeals court in Philadelphia will hear a case about a small t-shirt company. Over three years, that company sold about $23,000 worth of merchandise with Penn State logos on it. A jury said that was IP infringement and ordered the company to pay $28,000. Penn State is appealing anyway. The university told the court that the trial judge’s rulings cost it as much as $60 million.

Look at those two numbers again. The company made $23,000. The university says it should have been paid sixty million. That gap is the story here, and it points to a rule that the entire licensing business has relied on for fifty years without any court ever proving it was sound.

The issue in the case? Whether putting a school’s logo on merchandise is automatically treated as trademark infringement, or whether the school still has to prove that consumers are likely to be confused about the source of the trademark.

The question nobody wants to answer

Trademark law has one basic job. It stops people from being fooled about who made something. If you buy a pair of Nikes, the Swoosh tells you Nike made them. Someone who sticks a Swoosh on a knockoff sneaker is lying to you, under the law, and the law steps in.

Now think about a Penn State t-shirt.

Nobody standing in a parking lot in State College on a Saturday believes the university runs a t-shirt factory. Nobody buys a vintage Nittany Lions tee because they trust the school’s sewing standards. They buy it because they went there, or their dad did, or they think the old logo from 1968 looks cool. The logo on that shirt is not telling you who made it. It is telling everyone else who you are.

So is that a trademark at all, or is it just a picture?

How the law got around the problem

Courts solved this by making “confusion” mean something broader.

Under the modern rule, you count as confused in two situations. The first is obvious: you are wrong about who manufactured the shirt. The second is the one that matters: you wrongly believe the university gave its permission. The fan knows Penn State did not print the shirt. But he might assume the school signed off on it, and that assumption by itself is enough to make the sale illegal.

This started with a 1975 case called Boston Professional Hockey Association v. Dallas Cap & Emblem.

A company called Dallas Cap sold embroidered patches with NHL team logos on them. The patches were loose. You bought one and sewed it onto your own jacket. The trial court in Texas said this was fine. Nobody was fooled about who made the patch, the patch was just a decoration, and the hockey teams were not in the embroidery business and had no reputation there to protect.

The appeals court disagreed, and its reasoning is worth understanding because everything since has been built on it. The logo covered the entire patch. That meant the logo was the product. Nobody bought the patch for the thread. They bought it for the symbol. So the seller was using the team’s trademark to sell goods, and that was enough, whether or not a single customer had been misled about anything.

That idea now has a name. Lawyers call it the merchandising right. It means a brand can stop you from putting its logo on a shirt without having to prove anyone was confused.

Another court said no

Five years later, the Ninth Circuit refused to follow the merchandising rule in a case called Job’s Daughters v. Lindeburg. The judges made a simple point. People wear emblems to show what they belong to. Our clothes and our jewelry and our cars are covered with the names of schools we went to, places we visited, and teams we root for. It would be silly, the court said, to assume a customer wants the emblem because he thinks the organization manufactured the jewelry.

Even the Fifth Circuit backed off its own decision within two years, in a Kentucky Fried Chicken case where it said a trademark is not just property to be protected no matter what it does in the market.

And then the disagreement sat there. For forty five years. The Supreme Court never took the question. The circuits never worked it out.

What settled the issue was not a court. It was money. Universities built licensing programs. Penn State has run one for more than forty years, with an agent who reviews every company that wants a license. The pro leagues built bigger ones. Collegiate licensing grew into a business worth billions. Once everybody in an industry is paying for something, a court can look around and say that customers must expect payment to be happening. The rule made itself true by being followed.

The part everyone is missing

Here is what makes the Penn State case interesting.

The university asked the trial judge to apply the easy version of the rule, the one where printing a school’s name on a shirt is infringement almost by definition. Judge Matthew Brann said no. He pointed out that neither the Supreme Court nor his own appeals court had ever settled the question. He made Penn State prove actual confusion, the hard way, in front of a jury.

Penn State won anyway.

The jury heard about the company’s disclaimers, its own brand tags, and its argument that the images were old and decorative. It deliberated for a day and found willful infringement on every mark. It rejected both of the company’s main defenses.

That result should change how brands think about this. The shortcut was never the thing holding the licensing business together. Penn State proved its case without it. Brands that have built their whole enforcement strategy on the assumption that any logo use is automatically illegal are standing on something weaker than they believe.

Why $60 million

Penn State’s appeal is mostly about counterfeiting, and counterfeiting is where the real money in trademark law sits. If a court finds counterfeiting, the plaintiff can get triple damages, attorney fees, and a fixed dollar amount per mark per product category. That is how $23,000 in sales turns into a sixty million dollar demand.

The trial judge threw that claim out in early 2024. Penn State argues he read the counterfeiting law too narrowly, so narrowly that even an exact copy of licensed merchandise would not count.

If the appeals court agrees, the biggest effect will not land on companies like Vintage Brand. Those are rare, visible, and able to hire lawyers. It will land in the thousands of cases filed every year in certain federal courts against anonymous overseas sellers, where nobody shows up to argue the other side and the counterfeiting multiplier does all the work. A rule written with one print on demand company in mind gets applied within weeks to defendants who never appear.

That cuts both ways. A broad counterfeiting rule is a great tool if you own a real licensing program. It is a serious problem if you are a small brand accused of copying a logo you thought was generic, because triple damages and fee shifting end most disputes long before anyone sees a courtroom.

What I would tell a client

There is a second case worth watching. Baylor has signaled it is moving toward settling its lawsuit against Boston University over the interlocking BU logo. Baylor filed in the Fifth Circuit, where the easy version of the rule applies and its path was much smoother. It is settling anyway. That tells you how much of this fight is about the law and how much is about who can afford to test it.

But the practical lesson has nothing to do with the appeal.

A licensing program is evidence. Penn State did not win because the law handed it a freebie. It won because it could show a jury forty years of reviewing licensees, checking quality, and acting like an organization that stands behind the goods carrying its name. If you license casually, approve everyone who asks, and police nothing, you are not only losing revenue. You are building the record that gets used against you the first time you try to stop somebody.

The court hears both sides this week. Whatever it decides, someone is going to ask the Supreme Court to take the case. Fifty years is a long time for an industry this large to run on a question nobody ever finished answering.

The Oxford English Dictionary contains something on the order of six hundred thousand words, which sounds like…a lot! Set aside the obsolete, the technical, the regional, and the unpronounceable, and a fluent adult English speaker actually recognizes perhaps twenty to thirty-five thousand. Narrow that to words a reasonable person would consider putting on a bottle, a storefront, or a term sheet, and you are down to a few thousand. That is the real inventory. That’s the shelf!

In 2018, two New York University law professors, Barton Beebe (my former law school trademark professor) and Jeanne Fromer, decided to find out how much of that shelf was still empty. They pulled every trademark application filed at the United States Patent and Trademark Office between 1985 and 2016, roughly 6.7 million of them, added the 300,000 marks already registered in 1985, folded in every office action the agency issued over a fourteen year stretch and all 128 million domain names then registered in .com, and compared the whole apparatus against the most frequently used words in American English. Their finding, published in the Harvard Law Review under the title “Are We Running Out of Trademarks?”, was that just over 81 percent of the thousand most common words in American English were already registered as single-word marks, and that of the 86,408 most frequently used English words, 97.1 percent were either registered or close enough to a live registration to be treated as confusingly similar.

Sit with that second number. If you open a dictionary, point at a word an ordinary person would actually use, and try to build a company on it, the base rate says you will collide with somebody roughly nineteen times out of twenty.

The pressure has only intensified since. Applicants filed more than 824,000 new classes at the USPTO in fiscal year 2025, and the American register now holds something north of 3.6 million live registrations. Worldwide, according to the World Intellectual Property Organization, there are approximately 93.2 million active trademark registrations in force, of which China alone accounts for 49.8 million, more than half the planet’s total. Every one of those is a small fence around a piece of commercial language, and the fences have been going up faster every year.

This is the environment in which a founder sits down with a legal pad, or more often a whiteboard and an agency invoice, and picks a name.

Marks before there was law

It helps to know that this system is very old, and that it did not begin as a marketing apparatus.

Roman potters stamped their wares, and the good ones were counterfeited briskly enough that archaeologists still turn up knockoff lamps bearing the mark of the Fortis workshop across the empire. Medieval guilds required their members to apply marks, and the reason was not brand equity but accountability. Under the English bread assizes, a baker’s mark on a loaf meant that when the loaf came up short of the legal weight, the authorities knew whose door to knock on. The cutlers of Sheffield were assigned marks by their guild, marks that were recorded, inherited, and litigated over. The earliest function of a trademark was to answer the question of who was responsible for the thing you had just bought, which is worth remembering, because it is still the answer to why trademark law exists.

The English common law took its time catching up. There is an oft-cited seventeenth-century reference in *Southern v. How* to a clothier who was defrauded when another man copied his mark, mentioned almost in passing. Real doctrine arrived in the nineteenth century, when *Sykes v. Sykes* in 1824 allowed a shot-belt maker to recover against a competitor stamping goods with his mark, and *Millington v. Fox* in 1838 established in equity that a defendant could be enjoined even without proof that he meant to deceive anyone. Registration systems followed. Britain’s opened on the first business day of 1876, and the first mark on the register, No. 1, was the red triangle of Bass ale, secured by a brewery employee who is said to have waited outside the registry through the night.

The American path was less orderly. Congress passed a federal trademark act in 1870, and the Supreme Court struck it down in 1879 in the Trade Mark Cases, holding that Congress had improperly grounded the statute in the constitutional clause covering patents and copyrights, which speaks of authors and inventors and their discoveries, and trademarks are neither invented nor authored so much as adopted. Congress tried again in 1881, this time under the commerce power, and again in 1905, before finally producing the Lanham Act of 1946, which is still the statute we live under.

Two features of that inheritance generate nearly every naming disaster you have ever heard of.

The first is that in the United States, rights come from use, not from registration. The federal register is a powerful instrument, and I will spend the rest of this piece telling you to use it, but a business in Ohio that has been selling under a name since 2011 without ever filing anything has enforceable common-law rights in its trading area, and those rights are invisible to anyone searching only the government database. The register is a map of a territory it does not fully describe.

The second is that trademark rights are territorial and sectoral. They stop at national borders, and within a country they attach to particular goods and services. Delta is a faucet and Delta is an airline, and nobody is confused. The consequence, though, is that owning a name in the United States buys you nothing in Australia, or Brazil, or China, and this is where the expensive stories begin.

Four ways it all goes wrong

“You lose a country, permanently.”

In 1970, Jack Cowin acquired the Burger King franchise rights for Australia and discovered that the name was already registered there by Don Dervan, an American who had opened a drive-in in an Adelaide suburb in 1962 and was, by the end of the decade, selling more than a million burgers a year in South Australia. Burger King’s corporate parent handed Cowin a list of marks it already owned and told him to choose. He picked Hungry Jack, the name of a Pillsbury pancake mix, added an apostrophe and an s, and that is why one of the largest fast food chains on earth operates under a different name in an entire continent, more than fifty years later, with over four hundred locations. The trademark obstacle eventually evaporated, and by then the local brand equity was too valuable to abandon. One examiner search in 1970 would not have changed the outcome, but it would have changed the timeline and the cost, and the point stands: the map of your brand is drawn by other people’s earlier filings.

The pattern recurs everywhere once you start looking for it. Unilever’s Axe body spray sells as Lynx in the United Kingdom, Ireland, and Australia because the Axe name was spoken for. TJ Maxx is TK Maxx across Europe for the same reason. Anheuser-Busch and the Czech brewery Budějovický Budvar have been fighting over Budweiser in country after country for more than a century, with the result that in parts of Europe the American product sells simply as Bud while the Czech product sells in the United States as Czechvar. These are not obscure companies with careless lawyers. They are the largest consumer brands in the world, operating under permanent linguistic compromises because somebody else got to the word first in a jurisdiction that mattered.

“You buy your way out, at a price set by the other side.”

Apple wanted the iPad name and, sensibly, went about acquiring it. Through a British shell entity created for the purpose, it paid the Taiwanese arm of a display manufacturer called Proview roughly £35,000, about $55,000, for the rights in various countries. The problem was that the Chinese registration was held by a different Proview entity in Shenzhen, which had registered IPAD there in 2001 and which took the position that it had never sold anything. Chinese courts agreed. Proview sought to block iPad sales across the mainland and initially demanded figures in the hundreds of millions. In July 2012, with the product already a global phenomenon and China its second-largest retail market, Apple paid $60 million to end it. The delta between the price of the rights properly acquired and the price of the rights acquired under duress was roughly a thousand to one.

Meta paid the same figure, $60 million, in December 2021, to a Sioux Falls bank holding company called Meta Financial Group, using a Delaware entity named Beige Key LLC so that the buyer’s identity would not inflate the ask. Meta Financial took a year to phase the name out and re-emerged as Pathward. Facebook had begun those discussions before announcing the rebrand, which is the correct sequence, and it still paid eight figures for a word it had already decided to build a trillion-dollar identity around.

“You get enjoined in the middle of your own launch.”

In October 2011, Research In Motion announced that its next-generation operating system, on which the entire future of BlackBerry depended, would be called BBX. A software company in Albuquerque called Basis International had been selling developer tools under BBx since 1985. Basis wrote to RIM. RIM declined to stop. Basis went to federal court in New Mexico and, on December 6, obtained a temporary restraining order barring RIM from using the name at its own developer conference in Singapore, which began the next morning. The court observed that the products reached the same class of consumers through the same channels, was unmoved by the argument that a conference in Asia would not affect American developers, and noted that one could reasonably infer RIM had chosen the mark to benefit from goodwill it had not built. RIM announced the name BlackBerry 10 from the stage. A company in the most consequential product transition of its life spent that week rebranding under judicial compulsion, and every analyst covering it wrote the word “misstep.”

“You win too thoroughly and lose the word.”

The other direction of failure might be more elegant but it’s just as fatal. A mark that becomes the name of the thing itself stops functioning as an indicator of source and passes into the public domain, a fate the law calls genericide. Aspirin was Bayer’s. Escalator was Otis’s. Thermos, cellophane, yo-yo, zipper, and dry ice were all somebody’s property before they became ordinary nouns, and each was lost through a combination of the owner’s own promotional language and a failure to police how everyone else used it. This is what it looks like when a naming decision succeeds beyond its own capacity to survive.

What clearance actually involves

Most people believe they have searched for a name when they have typed it into the USPTO’s public database and Google, found nothing alarming, and bought the domain. I have watched a venture fund lose a name that way after deploying more than $40 million under it. They checked those two sources in 2021, built the thesis, the identity, and the website around the result, and received a cease-and-desist and then a lawsuit from a similarly named family office years later.

Here is what a real search covers, in plain terms.

Identical hits are the least of it. The legal standard is not identity but likelihood of confusion, assessed through a multi-factor inquiry that considers the similarity of the marks in appearance, sound, and meaning, the relatedness of the goods, the channels of trade, the sophistication of the buyers, and the strength of the earlier mark. That means a search has to reach phonetic equivalents and near-misses, the misspellings and vowel-drops that founders love, the singular and plural forms, and the translations, because a doctrine of foreign equivalents can make an English word confusable with its Spanish or French counterpart in the eyes of an examiner.

It has to reach beyond the federal register into state registrations, corporate name filings, domain registrations, social handles, app stores, industry directories, and ordinary commercial use, because of the common-law rights problem described above. It has to reach pending applications that have not yet published, including intent-to-use filings by companies that have not launched anything and may not for years. It has to account for the six-month priority window under the Paris Convention, which means somebody who filed in Germany five months ago can file here next week and claim a date that precedes yours. And if you have any ambition outside the United States, it has to reach the registers of the countries you intend to enter, before you enter them, because that is the Hungry Jack’s problem and it does not fix itself.

Then a lawyer has to read the results and form a judgment, which is the part no database performs. A search report is a list. Clearance is an opinion about which entries on that list are genuinely threatening, which are dormant registrations vulnerable to a cancellation petition for non-use, which sit in adjacent classes owned by companies with a documented appetite for opposing everything that moves, and which can be defused with a coexistence agreement or a narrowed description of goods. The output that matters is not the list. It is a recommendation about whether to proceed, and on what terms.

The stakes are not merely commercial, either. Beebe and Fromer point out that in pharmaceuticals, congestion in the naming space is a safety problem, with somewhere between 8 and 25 percent of medication errors attributed to confusion between similar drug names. There is a reason the FDA reviews proposed names separately from the trademark system.

Choosing a name that can be defended

The framework every American trademark lawyer carries in their head comes from Judge Henry Friendly’s 1976 opinion in *Abercrombie & Fitch Co. v. Hunting World*, which arranged marks along a spectrum of distinctiveness that has governed ever since.

Generic terms cannot be trademarks at all. You may not own Coffee for coffee, no matter how much you spend.

Descriptive terms, which tell the consumer what the product is or does, can be registered only after they acquire secondary meaning, which is to say only after years of use and expenditure have taught the public to hear the term as a brand rather than a description. Most founders want a descriptive name, because it does free explanatory work in a crowded market, and most founders are therefore choosing the weakest available option.

Suggestive terms require a small leap of imagination between the name and the product, and are protectable immediately.

Arbitrary terms are real words applied to unrelated goods, which is Apple for computers.

Fanciful terms are invented outright, which is Kodak, Xerox, Verizon, and every startup that dropped a vowel or bolted two syllables together.

The strength ranking runs in exactly the opposite direction from the marketing instinct. The names easiest to explain are hardest to own, and the names hardest to explain are easiest to defend. This is the whole reason the contemporary naming landscape looks the way it does, thick with coinages and portmanteaus and Latin fragments, and Beebe and Fromer found the shift in the data: as depletion increases, applicants file for longer marks, more syllables, more words, more invention, settling for what the authors call second-best options because the first-best ones are gone.

My practical advice, offered as someone who has cleaned up more of these than I have prevented, is unglamorous. Generate more candidates than you think you need, because you will lose most of them and you do not want to be negotiating from a shortlist of one. Run knockout searches on the whole set before anybody falls in love, since attachment is an adversary here and it forms fast. Do the full search and get the opinion before you commission the identity, buy the domain, print the packaging, or announce anything, because every one of those steps raises the cost of changing course and none of them raises your legal position. If international expansion is remotely plausible, search the destination markets while the name is still theoretical. And file, promptly, because in a system where 97 percent of the useful vocabulary is spoken for, priority is the primary durable asset in the room.

The register is public. Every abandoned application, every refusal, every opposition sits there in the record, permanently legible to anyone who cares to look, which means the mistakes are documented in a way that few other business errors are. There is no reason to add to the archive when the alternative is a search that costs less than a week of the agency retainer you are about to spend on the logo.

Here’s something a lot of people don’t realize: what clients buy from lawyers has never really been hours. Hours are how we have priced the work for the last 100 years or so, which is a much shorter run than most people assume. What clients actually buy is a degree of confidence that a problem has been handled, and the interesting question raised by AI is not whether we can produce documents faster. Of course we can. The question is whether anyone is prepared to buy a different level of confidence at a lower price, and whether the profession has the vocabulary to describe what that would even mean.

In other words, are clients ready to pay lower fees in exchange for foregoing human review of AI work product for certain types of work?

That’s the real question to ask. Read on, I’ll explain.

The billable hour is a recent invention

The billable hour emerged in American law through a side door, and it was more about management than profit. Reginald Heber Smith, a young Harvard graduate who spent five years as counsel to the Boston Legal Aid Society before becoming managing partner of Hale and Dorr in 1919, is generally credited with introducing systematic timekeeping to law practice (damn you Reggie!). His 1919 book, “Justice and the Poor”, became a foundational text in the legal aid movement, and the organizational zeal he brought to legal aid came with him to the firm. His timesheet was, at first, a purely internal instrument of planning and budgeting, and his partners resented it as an indignity borrowed from the factory floor, a stopwatch pointed at gentlemen.

For decades the timesheet stayed inside the firm. Prices came from somewhere else, namely the minimum fee schedules published by bar associations, which set what lawyers could charge for a will or a title search.

Two things then converted an internal metric into the profession’s price. The first was a campaign by the American Bar Association’s Special Committee on Economics of Law Practice, whose 1958 pamphlet “The 1958 Lawyer and His 1938 Dollar” argued that lawyers were poor businessmen who failed to record their time, and calculated that a lawyer had only about thirteen hundred fee-earning hours in a year. The second was the collapse of the alternative. In Goldfarb v. Virginia State Bar, decided in June 1975, the Supreme Court held that the Fairfax County minimum fee schedule, enforced by the Virginia State Bar through the threat of discipline against lawyers who charged less, was price fixing under the Sherman Act, and that there was no sweeping exemption for the learned professions. By the late 1970s, hourly billing was the dominant method of pricing legal work in the United States.

In other words, flat fees themselves were never illegal; what Goldfarb struck down was the bar associations' collectively published minimum fee schedule, enforced through the threat of discipline, which was horizontal price fixing under Section 1 of the Sherman Act.

Think about that - - hourly billing became the price of legal services not because anyone concluded that time was a good measure of legal value, but because a management tool was already sitting on every lawyer’s desk when the old pricing convention became illegal. It has been about fifty years. Institutions of that vintage feel permanent from the inside, and they are not.

We’ve priced assurance before

There is an older episode that speaks more directly to the question of whether clients would accept less human review in exchange for a smaller bill, and it is a story about lawyers being displaced not by a machine but by a financial product.

In February 1868 the Pennsylvania Supreme Court decided Watson v. Muirhead. Muirhead, a conveyancer, had searched the title on a Philadelphia property for Watson, the purchaser. He found recorded judgments, took the question to a lawyer, and was advised that they were not valid encumbrances. The advice was wrong. Watson bought the property and was left to satisfy the liens himself. When he sued, Justice Sharswood held for the conveyancer, reasoning that a professional’s conduct is not to be judged by the result, and that a conveyancer, like a lawyer or a physician, should be protected when he acts to the best of his skill and knowledge. Watson, who had done nothing wrong and had paid for exactly the professional care the system offered, absorbed the entire loss.

The response was not to demand more careful title searches. In 1874 the Pennsylvania legislature authorized the incorporation of title insurance companies, and in 1876 a group in Philadelphia formed what is generally regarded as the first of them. What that industry sold was not a better opinion. It was a transfer of risk to a capitalized institution, priced as a premium and payable on a defined set of losses whether or not anyone had been negligent. The human search did not disappear, and it continues today inside every title plant in the country, but the buyer stopped paying for a professional’s judgment and started paying for a promise with a number attached to it.

A century later, the Goldfarbs went looking for a lawyer to examine the title on their house in Fairfax County only because their lender required them to buy title insurance. The insurance product created by the failure of a professional opinion generated the transaction that eventually destroyed the profession’s price list.

The lesson in that sequence is counterintuitive. The market did not resolve the problem by accepting sloppier work at a discount. It resolved it by making the risk explicit, pricing it, and moving it onto a balance sheet that could absorb it.

Accountants built a vocabulary for this. Lawyers did not.

The comparison that keeps nagging at me is with auditing, because auditors faced a version of the efficiency question a century ago and answered it in a way that reshaped their profession.

Nineteenth century audits aspired to complete examination. The auditor looked at every transaction, and the value of the audit was understood to rest on that completeness. As companies grew, complete examination became impractical, and the profession moved to test checking and eventually to formal statistical sampling. What made that shift survivable was not the sampling technique itself but the language that grew up around it. Auditors developed the concepts of materiality and reasonable assurance, which are, stripped of jargon, an agreement between the auditor and the world about how much confidence is being purchased and what falls outside the promise. An audit opinion is a carefully bounded statement, and everyone who relies on it knows it is bounded.

Law has almost nothing equivalent. A legal opinion letter comes closest, with its stated assumptions and qualifications, and it is telling that opinion practice is the corner of legal work where lawyers are most precise about what they are and are not vouching for. Everywhere else, a client who asks how confident we are receives an answer in adjectives. We say a position is strong, or defensible, or aggressive. We do not say that we reviewed a sample sufficient to reduce the probability of a material error below some stated threshold, because we have never had to, and because our fee has always been quoted in a unit that describes effort rather than confidence.

That absence is why the AI billing conversation keeps going in circles. A client asking for an AI discount is asking to buy less of something, and neither side has a name for the thing.

Law’s one completed experiment

There is one area where the profession ran this experiment to conclusion, and it did not go the way the folklore suggests.

In 1985, David Blair and M. E. Maron published a study of a full text retrieval system used in the defense of litigation arising from a Bay Area Rapid Transit accident. Lawyers and paralegals ran and revised queries against a collection of about forty thousand documents, roughly three hundred and fifty thousand pages, working from fifty-one information requests and continuing until they were satisfied they had found at least seventy-five percent of the relevant material. Measured against the full collection, they had found about twenty percent. Precision was high, around eighty percent, which is part of the point. The documents they found looked right. The finding that mattered was that the searchers had no way to gauge what they were missing, because the missing documents used language nobody had thought to look for.

That study became the empirical foundation for a change in practice a quarter century later. In February 2012, Magistrate Judge Andrew Peck of the Southern District of New York approved the use of predictive coding in Da Silva Moore v. Publicis Groupe, the first published decision to do so. His reasoning is the part worth remembering. He did not hold that the software was excellent. He held that the standard against which it should be judged was not perfection but the realistically achievable alternative, observed that the belief in manual review as the gold standard was a myth rather than a finding, and framed the whole question in terms of proportionality, with the goal being higher recall and precision than the alternative method at a cost sensible relative to the value of the case. Maura Grossman and Gordon Cormack had published research in the Richmond Journal of Law and Technology the year before showing that technology assisted review could outperform exhaustive human review on both measures.

Two lessons follow, and they curiously point in opposite directions.

The first is that the machine was accepted only once the profession agreed on a way to measure the answer. Recall and precision are unglamorous statistics, but they gave courts and parties a shared language for how good a review was, which meant the tradeoff could be argued about rather than merely asserted. The second lesson is that discovery did not get cheaper in the way anyone promised. Document volumes grew faster than the tools improved, and much of the savings was absorbed by scope. The efficiency was real, and it went into doing more, which is what usually happens when the cost of an activity falls.

Where the rules are on this issue

For lawyers billing by the hour, part of the question has already been answered, and not by the market. ABA Formal Opinion 93-379 established that a lawyer who has agreed to bill hourly cannot charge for more time than was actually spent, and that a lawyer who turns out to be particularly efficient at accomplishing a result does not get to keep the difference. Formal Opinion 512, issued on July 29, 2024, applies that principle directly to generative AI. The opinion’s own example is a lawyer who spends fifteen minutes feeding information into a tool to draft a pleading and may bill for those fifteen minutes plus the time actually spent reviewing the draft for accuracy and completeness. Lawyers may not bill clients for the time they spend learning to use tools they will use across their practice, though they may charge for learning a tool the client specifically asked them to use, if that is agreed in advance. The opinion also cautions that a flat fee can itself become unreasonable under Model Rule 1.5 if the work behind it has collapsed.

So under hourly billing, the efficiency gain belongs to the client by operation of the ethics rules. The pressure this creates is obvious, and it is a large part of why flat and fixed fees are moving from the margins toward the center. A flat fee is not merely a pricing convenience. It is a transfer of risk, because the lawyer who quotes a fixed number has agreed to absorb the cost of being wrong about how long the work will take. That is closer to the title insurance model than to the timesheet, and it is the direction the economics push once the hour stops describing the work.

The reason review is not going away soon

None of this suggests that verification is a superstition the profession will grow out of. The public record on that is now substantial. Damien Charlotin, a research fellow at HEC Paris, maintains a database of decisions in which a court found, or clearly implied, that a party relied on material fabricated by generative AI. As of its September 5 update it listed 2,022 cases, up from roughly two hundred in the middle of last year, with 1,379 from courts in the United States. The sanctions have escalated from the five thousand dollars imposed on two New York lawyers and their firm in Mata v. Avianca in June 2023, where a brief cited six opinions that did not exist, to six figure awards and, in a small number of matters, suspension.

Two caveats keep that number honest. The majority of the entries involve self-represented litigants rather than lawyers, with the database attributing roughly 1,163 to pro se filers and 805 to lawyers. And the database counts only what a court noticed and wrote up, so it is a floor rather than a measure.

The tools built for lawyers are better than a general purpose chatbot, but they are not nearly perfect. A Stanford RegLab study by Varun Magesh and colleagues, published in the Journal of Empirical Legal Studies in 2025, tested the leading legal research products and found hallucination rates between seventeen and thirty-three percent, against forty-three percent for GPT-4. The testing was done in 2024 and the products have been revised since, so the specific numbers are a snapshot rather than a verdict. The finding that survives is the structural one, which is that retrieval grounding reduces fabrication without eliminating it, and that a longer, more useful answer contains more propositions that can be wrong and takes more time to check.

The current standard of care, then, is the one my original post described: use the tools, and verify everything that leaves the building. The floor of human work this creates is real, and it is a large part of why bills have not fallen in proportion to the speed of drafting.

What is less often noticed is that this floor is not fixed by nature. It is a judgment about custom, and custom in the law of negligence has never been the last word. In The T.J. Hooper, decided by the Second Circuit in July 1932, two tugs towing coal barges from Norfolk to New York lost a barge each in an easterly gale off the Jersey coast. The weather bureau at Arlington broadcast two forecasts a day, and the evening forecast, which the masters never heard, predicted the worsening conditions that would have sent them into the Delaware Breakwater for shelter. The detail that makes the case hit differently today is that both masters did own radio receiving sets, kept aboard as personal property, partly toys, neither supplied by the owner nor maintained by anyone. They were not in working order. Learned Hand found no general custom among coastwise carriers either way, held the tugs unseaworthy regardless, and wrote that an entire calling may lag in adopting new and available devices, that a trade may not set its own standard of care however settled its habits, and that courts must in the end say what is required.

The parallel is not that lawyers must adopt AI. It is that a tool sitting on the desk, personally acquired, unsupervised by the firm and not integrated into any system of verification, is roughly the posture the Second Circuit found insufficient in 1932. The question today is whether a lawyer adequately checked the machine. The question in a decade may be whether a lawyer who declined to use one, and therefore missed something a competent tool would have caught, met the standard of care.

The conversation worth having

The people least served by all of this are the ones who never appear in the debate. The Legal Services Corporation’s 2022 study found that low-income Americans received no legal help or inadequate legal help for ninety-two percent of the civil legal problems that affected them substantially, that they sought help for only one problem in four, and that among those who did not seek help, forty-six percent gave cost as a reason. Households with higher incomes did somewhat better and not dramatically so. Those people are not making a considered judgment about assurance levels. They are looking at the price of the only product on offer and deciding to take their chances with a form, a template, or nothing at all. That is the tradeoff at the heart of the AI billing question, made silently and at enormous scale, every day. It is also close to the problem Reginald Heber Smith set out to address when he handed his partners a timesheet, which is a symmetry worth noticing, because the instrument he built to make legal work measurable is the one that eventually made it expensive.

For clients and lawyers who are in the room together, the conversation that would actually move things forward is narrower and more concrete than the one we are having. It starts by separating work where an error is recoverable from work where it is not, since a first draft of a routine vendor agreement and a representation letter in a financing carry entirely different tails. It continues by naming the level of verification being purchased, in terms specific enough to write down. And it ends, if it goes well, with a price that reflects an agreed allocation of risk rather than a count of hours.

That is not a technology conversation. It is the same one Pennsylvania had after *Watson v. Muirhead*, and the accounting profession had when complete examination became impractical, and the litigation bar had when the document collections outgrew the reviewers. In each case the resolution came from being explicit about what was being promised, and in each case it took a generation. We are overdue, and the tools have made the delay harder to ignore.

What we are doing at the moment is asking for AI pricing while continuing to require human levels of assurance, and the historical record says those two things are reconciled by identifying the risk and pricing it rather than by discounting the review.

That conversation has not started in most engagements, which is the actual reason the bills look the way they do (largely the same!).

---

*Sources: WilmerHale firm history and Thomson Reuters Institute on Reginald Heber Smith and the origins of legal timekeeping; ABA Special Committee on Economics of Law Practice, The 1958 Lawyer and His 1938 Dollar (1958); Goldfarb v. Virginia State Bar, 421 U.S. 773 (1975); Watson v. Muirhead, 57 Pa. 161 (1868), and the 1874 Pennsylvania act authorizing title insurance companies; Derek Matthews, A History of Auditing (2006); David C. Blair and M. E. Maron, An Evaluation of Retrieval Effectiveness for a Full-Text Document-Retrieval System, 28 Comm. ACM 289 (1985); Maura R. Grossman and Gordon V. Cormack, Technology-Assisted Review in E-Discovery Can Be More Effective and More Efficient Than Exhaustive Manual Review, 17 Rich. J.L. & Tech. 11 (2011); Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012); ABA Formal Opinions 93-379 (1993) and 512 (2024); Mata v. Avianca, No. 22-cv-1461 (S.D.N.Y. June 22, 2023); Damien Charlotin, AI Hallucination Cases database (updated September 5, 2026); Varun Magesh et al., Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools, 22 J. Empirical Legal Stud. 216 (2025); The T.J. Hooper, 60 F.2d 737 (2d Cir. 1932); Legal Services Corporation, The Justice Gap: The Unmet Civil Legal Needs of Low-income Americans (2022).*

I should say at the outset that I represent both sides, and that this is not a piece about how brands are predatory. Some are, sure. But most are simply companies with a fiduciary obligation to protect ultra-valuable assets. Here’s the story of how these deals work, and how we got to where we are today.

The ancestors

In 1902, a young woman named Abigail Roberson found her portrait on twenty five thousand flour advertisements and sued. The New York Court of Appeals found that no such right existed, which is why the legislature passed what became sections 50 and 51 of the Civil Rights Law the following year.

The modern concept arrived in 1953, and it arrived, appropriately enough, through a fight over chewing gum. Two confectioners had each signed the same ballplayers to exclusive card rights, and when the case reached the Second Circuit, Judge Jerome Frank observed that a prominent person has an interest in the commercial value of his likeness that can be assigned to somebody else, and gave it a name: the right of publicity. Every collaboration deal you will ever sign is downstream of that sentence. What Frank recognized was not merely a right to object but a right to convey, which is the thing that makes a deal possible and also the thing that makes a bad deal permanent.

The fee era

For the next quarter century, the shape of the bargain was simple and, from where we sit now, remarkably cheap. A company paid a flat fee for a period of use, the talent showed up, and the relationship ended when the campaign did. Arnold Palmer and Mark McCormack shook hands in 1960 and built the modern representation business on the insight that an athlete’s commercial life could be managed as a portfolio rather than as a series of favors, and the agency they created professionalized the fee but did not really change its logic. Michael Jackson’s arrangement with Pepsi in 1983 was reportedly worth five million dollars, an enormous number at the time, and it was still, structurally, a rental.

The problem with rentals is that they end. The company keeps the goodwill that the campaign generated, the talent keeps the check, and if the campaign works spectacularly well, the talent has no participation in the spectacle. Everyone in the business understood this by the early eighties. What nobody had done was fix it.

Participation

The fix, when it came, came from a shoe company that was losing to Converse and had very little to lose. The Nike agreement with Michael Jordan in 1984 is famous for its size, but the size is not the interesting part. The interesting part is that Jordan received a royalty on sales of a product line that carried his name, which meant that his upside was tied to performance in the market rather than to the negotiating leverage he happened to have at signing. That is the single most consequential structural move in the history of this field, and its consequences are still unfolding. Nearly thirty years later, when Kanye West asked Nike for a royalty on Yeezy footwear and was refused, he went to Adidas, which said yes, and the reason that conversation was even possible is that Jordan had made the royalty the default expectation for anyone with genuine cultural weight.

Two years after the Jordan deal, Run DMC performed “My Adidas” at Madison Square Garden, asked the crowd to hold up their shoes, and produced a moment that persuaded a German footwear company to write a contract reportedly worth one and a half million dollars to a rap group. That deal mattered because it established that cultural authority, unattached to athletic performance, moved product on its own. Everything that the fashion industry now calls a “collab” probably descends from that sweaty night at MSG.

The counterweight arrived in 1989. Pepsi built a campaign around Madonna, the accompanying video for “Like a Prayer” generated a religious backlash, and the company withdrew the commercial while Madonna kept her fee. Read that outcome carefully, because it is the origin of an entire genre of contract drafting. Brands ultimately concluded that they needed morals clauses with teeth, and talent concluded that fees should be paid on execution rather than on airing. Both sides were right, and both provisions are in every agreement I see today.

The courts were busy in the same period, and their work has aged into unexpected relevance. When Bette Midler declined to license a song to Ford, the agency hired a singer who could imitate her, and the Ninth Circuit held in 1988 that deliberately imitating a distinctive voice to sell a product is actionable. Tom Waits won a jury verdict of about two and a half million dollars against the snack company behind Doritos on similar reasoning in 1990. Around the same time, Vanna White obtained a favorable ruling against Samsung over a print advertisement featuring a robot in a blonde wig standing beside a letter board, over a dissent from Judge Kozinski warning that the right of publicity was expanding into territory where it would start eating expression. Those cases were about impersonation by human beings and analog machinery. They are now the closest thing we have to precedent for what a generative model does.

Ownership

By the 2000s, the sophisticated version of the bargain had moved again, from participation in revenue to participation in the enterprise. Curtis Jackson took equity in Glacéau rather than a larger fee, and when Coca Cola acquired the company in 2007 he reportedly cleared something in the neighborhood of a hundred million dollars before taxes. Dr. Dre and Jimmy Iovine built a headphone company and sold it to Apple in 2014 for three billion. George Clooney and his partners sold Casamigos to Diageo in 2017 in a transaction reported at up to a billion dollars, and Ryan Reynolds followed a similar path with Aviation Gin three years later. None of these were endorsements. They were businesses in which a famous person’s involvement was the marketing budget.

The fashion houses reached the same conclusion by a different route. Louis Vuitton sent a nastygram to Supreme in 2000 over a monogram treatment on skateboards and then, seventeen years later, produced a collection with the same company, which tells you almost everything about how the luxury sector metabolizes the culture it once litigated against. Virgil Abloh’s reconstruction of ten Nike silhouettes in 2017 moved the designer from hired hand to author. Rihanna’s partnership with LVMH in 2019 made her the operating principal of a maison rather than its face, and although the ready to wear venture closed in 2021, the beauty and lingerie businesses did what almost no celebrity brand does, which is to sustain revenue that does not depend on the founder’s personal appearances.

And then there is the deal that everyone in my position now uses as the cautionary example. When Adidas terminated its relationship with Ye in 2022, it discovered that it could not simply keep selling. He owned the trademark. Adidas owned the designs and the inventory, roughly a billion dollars of which sat in warehouses. Neither party could go forward alone, which is the worst possible position for both, and it was the predictable result of a structure in which the intellectual property had been divided along a line that made sense during the romance and none at all during the divorce. If you take one thing from the history, take this: the question of who owns what at the end is more important than the royalty rate at the beginning, and it is almost always negotiated with less attention.

Disclosure

Running alongside the deal architecture is a regulatory story that most creators still underestimate. The Federal Trade Commission has policed endorsements for decades, but the modern posture dates to the 2009 revision of its Endorsement Guides and the enforcement that followed. Lord and Taylor settled in 2016 over a campaign in which fifty influencers posted the same dress without disclosing that they had been paid. Warner Brothers settled the same year over sponsored gameplay videos. In 2017 the Commission sent warning letters to roughly ninety influencers and brands, which was the moment the industry understood that the individual, not only the company, was in the frame. Kim Kardashian paid about 1.26 million dollars to the Securities and Exchange Commission in 2022 over a single promotional post concerning a crypto asset. The Commission updated the Guides again in 2023 and finalized a rule addressing deceptive reviews and testimonials in 2024. The practical upshot is that disclosure obligations attach to the person with the audience, indemnity provisions in the contract may or may not save you, and a brand that drafts well will make you responsible for your own compliance.

What is actually on the table now

The live issues in the deals crossing my desk this year cluster into a handful of areas, and I want to describe them in a way that illuminates the concerns of each side in collabs.

The largest is synthetic likeness. Brands increasingly ask for rights to your image, voice, and performance in perpetuity, in all media now known or later devised, and, in the newer templates, for purposes of training and generating outputs with machine learning systems. The brand’s argument is legitimate on its face: production is expensive, localization requires generating dozens of variants, and it does not want to renegotiate every time it needs a new asset. The artist’s exposure is that a training grant is not a use grant. A campaign ends. A model does not. Tennessee’s ELVIS Act in 2024, California’s legislation on digital replicas the same year, and the federal proposal that would create a national right against unauthorized digital replicas all reflect a legislative judgment that consent to appear is not consent to be simulated, and the sensible contractual position is to license specific generated outputs for specific campaigns with approval rights, rather than to license the underlying capability.

The second is scope. Exclusivity clauses are frequently drafted by category, and categories have a way of expanding to swallow a career. A footwear exclusive that is defined as “athletic and lifestyle apparel and accessories” will prevent an artist from taking an eyewear deal, and the brand did not necessarily intend that outcome; it simply used a template. Negotiate the definition, not the concept.

The third is the ownership question that Adidas and Ye illustrate. Who owns the design, who owns any new mark created for the collaboration, who owns the social account if one is created, who owns the customer data, and what happens to unsold inventory when the relationship ends. A brand needs a period to sell through what it has manufactured, which is reasonable, and an artist needs that period to be finite and to carry restrictions on discounting and channel, which is equally reasonable.

The fourth is that morals clauses have become bidirectional, and should be. If the brand can terminate you for conduct that damages its reputation, you should be able to terminate the brand for conduct that damages yours, particularly where the artist’s public identity is bound up with positions the company may not share.

The fifth is that equity, which has become the fashionable ask, is not automatically better than cash. Equity in an operating business with real distribution can be transformative, as the beverage examples suggest. Equity in an entity that exists to hold one collaboration, subject to dilution, with no information rights and no protective provisions, is frequently worth less than the fee it replaced. Ask what you are getting a piece of before you ask for a piece. If you don’t know how preferred stock works upon a liquidity event, talk to someone who does before recommending that your client take equity in any deal.

What has changed over a hundred and twenty years is not the fundamental transaction, which remains an exchange of attention for money, but the sophistication with which the terms of that exchange are set, and the sophistication has accumulated much faster on one side of the table than the other. That gap is closeable. It closes when the person being asked to sign understands that the template in front of them is not a neutral description of an obvious arrangement but an accumulated record of every dispute the company has previously lost, and treats it accordingly, which usually means asking for less than they fear and getting more than they expect.

Somebody in marketing searches the company’s own product on Amazon and finds four hundred listings that are not yours. The photographs are your photographs, lifted from your campaign. The price is forty percent of yours. The reviews mention that the stitching came apart, and those reviews are attaching themselves to your name in the minds of people who will never know they bought a fake.

So you send takedown notices, which is what the platforms tell you to do, and it works, and within nine days the same operator is back under three new storefront names selling the same goods from the same warehouse. You do it again. They come back again. Somewhere in the second or third cycle of this your general counsel says the obvious thing, which is that we should sue somebody, and then discovers that there is nobody to sue. The seller is a name invented last Tuesday. The address is a mail drop. There are no books, no bank you can reach, and no realistic prospect of collecting a dollar even if you win. Whack-A-Mole.

Let’s talk about Schedule A. The freezing of defendants’ Paypal and other payment accounts is really the point, because it captures the seller's money where it sits, and that is the only way a brand ever actually recovers a dollar from a counterfeiter it could not otherwise find, serve, or collect from. .It is the most effective anti-counterfeiting tool American brands have ever had. A federal judge in Chicago spent last summer explaining, at length and persuasively, why it should not exist in its present form. Both of those things are true, and the second one is coming for the first, so it is worth understanding how we got here.

Why the ordinary remedies fail

The traditional remedy for trademark infringement is an accounting of the infringer’s profits, and it is a genuinely elegant instrument with a century of American law behind it. The Supreme Court blessed it in 1916, Frankfurter gave it its best statement in 1942, and Congress wrote it into section 1117(a) of the Lanham Act in 1946. Once a plaintiff proves the infringer’s sales, the burden shifts to the infringer to prove which portion of its profit came from something other than the stolen mark. It is a beautiful allocation of risk, and it presumes a defendant with a general ledger.

Against a counterfeiter it does nothing. There is no ledger to produce, no comptroller to depose, no bank account within reach of American process, and often no entity that will still exist by the time a judgment issues. You can win completely and recover zero. We’ve seen this happen hundreds of times over the years in our work here at Jayaram.

Congress recognized this in 1996 and passed the Anti-counterfeiting Consumer Protection Act, which added section 1117(c). A brand facing a counterfeiter may skip the accounting entirely and elect statutory damages instead, between $1,000 and $200,000 per counterfeit mark per type of goods, rising to $2 million per mark where the counterfeiting was willful. No proof of the defendant’s sales is required, because the defendant has arranged to make that proof impossible, and Congress placed the cost of that impossibility on the party who engineered it.

That was the substantive half. The procedural half was invented twenty years later, and not by Congress.

What the Northern District of Illinois built

Beginning in the early 2010s, brand-side lawyers filing in the Northern District of Illinois assembled a procedure that Congress never considered and that no appellate court has squarely reviewed.

A brand files a single complaint naming hundreds of online sellers at once, joined together under Rule 20 on the theory that they are all selling counterfeits of the same mark. The sellers are not named in the caption. They are enumerated on an exhibit filed under seal, which is where the practice gets its name. The brand then moves, without notice to anyone, for a temporary restraining order that does three things at once: it orders the listings taken down, it orders the marketplaces and payment processors to freeze the sellers’ accounts, and it authorizes service of process by email.

The first a seller learns of any of it is when the money stops.

The district embraced this with remarkable enthusiasm. Roughly 4,207 Schedule A cases were filed there between January 2013 and February 2025, and something on the order of eighty percent of all such suits nationally are filed in that one courthouse. At least two judges published fill-in templates for the restraining orders, preliminary injunctions, and default judgments. One judge, in a 2022 opinion, described the phenomenon as the Northern District of Illinois against the internet, which was meant wryly and turned out to be accurate.

The plaintiffs are not fringe operators. Volkswagen Group of America has filed there. So has Liverpool Football Club. So have European automotive brands, entertainment rights holders, footwear companies, and increasingly design patent owners, whose Schedule A filings in that district went from 38 in 2020 to 54 in 2021 to 63 in 2022.

What it actually delivers, which is not what the headlines say

Here’s the rub:

The eight-figure default judgments that occasionally reach the trade press are almost entirely worthless. A two million dollar judgment against a seller who does not exist, in a jurisdiction whose courts will not enforce it, is a piece of paper. Nobody in this practice is collecting on those, and any lawyer who tells you the judgment is the win is selling you something.

There’s four reasons why Schedule A is so effective.

The first is the frozen money, which is the only real recovery in the entire process. When the order lands, whatever the seller holds in its marketplace and payment accounts is captured before it can move offshore. Per defendant it is rarely enormous. Across two hundred defendants it becomes a number that funds the program and then some, and critically it is money you could not have reached by any other route.

The second is the listings, removed at a scale and speed that notice-and-takedown cannot approach. A court order binds the marketplaces as persons in active concert, so compliance is not a matter of platform goodwill or ticket queues. It happens within days, across every platform named.

The third is the domains and storefronts, which can be transferred or disabled outright rather than merely emptied. That is what breaks the reappear-under-a-new-name cycle that makes ordinary enforcement futile.

The fourth is deterrence you can actually observe. Operators talk to one another. A brand that runs one of these programs consistently develops a reputation in the seller communities, and the listings migrate toward brands that do not.

I run these programs. I think they work. That is the honest starting position from which everything below should be read.

Then the judges started reading their own dockets

And then, the pushback!

In December 2023, Judge Steven Seeger refused to enter a restraining order against 310 Schedule A defendants in a case brought by Zorro Productions, writing that what the plaintiff wanted was to strike at the defendants under cover of darkness. That was one judge in one case, and it was noticed.

In June 2025, Judge John Kness went considerably further. He stayed every Schedule A case on his docket, roughly fourteen of them, not because a defendant had appeared to complain, since in the ordinary course no defendant appears at all, but because he had decided to reconsider what he had been signing several times a week for years. In August he issued a comprehensive opinion in a case brought by an Indian motorcycle manufacturer, concluding that the model should not be perpetuated in its present form, and citing Eric Goldman’s 2023 law review article on the practice something like fifteen times. Goldman’s count, which the opinion adopts, is more than 7,700 cases affecting more than 1.5 million defendants.

The critique is not about the money provision. This is the distinction nearly every summary of the opinion collapses, and it is the one that determines whether the practice is reformable.

Statutory damages under section 1117(c) are fine. Congress enacted them deliberately, for stated reasons, and no judge has suggested otherwise. The problem is four procedural devices stacked on one another, each defensible alone.

Joinder is the first. Rule 20 permits defendants to be sued together when the claims arise from the same transaction or series of transactions, and hundreds of unrelated sellers who have never met do not obviously qualify. Congress foreclosed exactly this in patent cases in 2011, enacting section 299 to break up mass filings against unrelated defendants. There is no equivalent in the Lanham Act, and that absence is the entire reason this practice lives in trademark.

Sealing is the second, and its justification expires. Concealing the defendant list for a few days to prevent asset flight has a logic. Keeping it sealed afterward serves mainly to keep the practice out of view.

The ex parte order is the third. Rule 65(b) permits relief without notice only on specific facts showing immediate irreparable injury, plus an attorney’s certification of the efforts made to give notice. Judge Kness observed that he had not seen a filing making that showing as to each defendant, and that the practice inverts the rule by presuming every defendant will behave badly unless shown otherwise.

The asset freeze is the fourth, and it is where the stack becomes something none of its parts authorize.

The seam

In Grupo Mexicano de Desarrollo, S.A. v. Alliance Bond Fund, Inc., 527 U.S. 308 (1999), the Supreme Court held that a federal court in a purely legal action for money damages has no power to freeze a defendant’s assets before judgment. That authority did not exist in the English Court of Chancery in 1789 and so it does not exist now. Statutory damages are a legal remedy. On its face, Grupo Mexicano forecloses the freeze.

The workaround is to plead an equitable accounting alongside the statutory damages claim, which unlocks the freeze as ancillary to equitable relief, and then to elect statutory damages at default judgment. Judge Kness put his finger precisely on this. He noted that he had not seen a Schedule A plaintiff actually pursue the equitable monetary remedy at the end of a case. The equitable claim opens the door and is left behind in the lock.

That is the flaw. Not the damages, not the enforcement objective, not even the joinder standing alone. It is a legal remedy secured through an equitable device by pleading a claim nobody intends to pursue, with the freeze arriving before the seller knows a case exists.

What a reformed version looks like

None of the repairs requires a brand to give up enforcement, which is why I think the practice will survive in some form.

Sever the freeze from the damages theory, so that a plaintiff intending to elect statutory damages says so at the outset and does not obtain a prejudgment restraint on the strength of a claim it will abandon. If the freeze is genuinely necessary, Congress can authorize it directly, as it did for ex parte seizure of counterfeit goods in section 1116(d), with the findings and bond requirements that provision carries.

Unseal on a schedule, once the restraint is in place and the flight risk has passed.

Require per-defendant specificity, which Rule 65(b) already demands and which a judge reviewing several hundred screenshots cannot supply on his own.

Enact a section 299 analogue for trademark. This is the cleanest available fix, it has a legislative precedent that worked, and it would end mass joinder without touching any brand’s ability to sue any actual counterfeiter.

Where this leaves brands

Judge Kness did something unusual at the end of his opinion, which was to invite the plaintiff to seek certification for an interlocutory appeal and to say plainly that his district needed appellate guidance. The plaintiff had no reason to accept and, as far as I can tell, has not. So there is no vehicle, no Seventh Circuit ruling, and a written opinion of real persuasive force circulating among judges who have been signing these orders for a decade. The District of New Jersey has adopted a standing order. Individual judges have begun refusing joinder. None of that is precedent, and none of it is uniform.

What that means operationally, if you are running or funding one of these programs, is that the tool still works and the ground underneath it is moving. Which judge you draw now materially changes what you get, and the pleading practice that was standard in 2022 will get a case thrown out in front of some judges in 2026. The programs that keep working will be the ones that stop treating the mechanism as a form to fill in.

The strange part is that the machinery is now widely understood to be procedurally indefensible by many of the people operating it, and it continues to operate, because the only parties with standing to challenge it are anonymous, frozen, and better off settling than fighting. That is not a stable arrangement. It is not at all obvious what breaks it first.

As a matter of black-letter copyright law, the result was not indefensible. Someone had copied a protected sound recording without a license. What the opinion contains no trace of is any understanding of what sampling was, how long it had been going on, whether the practice had developed norms of its own, whether the defendants’ conduct was aberrant or entirely standard within their field, or what a rule of strict liability would do to a form of music that was by then more than a decade old and commercially ascendant. The judge did not think he needed that information, because he had framed the question as one of theft, and theft is a question you can answer without knowing anything about the thief’s profession.

The consequences ran for thirty years. Clearance became mandatory and expensive, which meant that records built from dozens of small fragments became economically impossible; the Bomb Squad’s production on It Takes a Nation of Millions to Hold Us Back and the Dust Brothers’ work on Paul’s Boutique belong to a window that closed. De La Soul’s catalog, which had been cleared under a licensing regime that never contemplated digital distribution, stayed off streaming services until March 2023. An entire aesthetic was priced out of existence by a decision that took eleven pages.

The point is not that Duffy reached the wrong result. Reasonable judges could reach that result. The point is that he reached a result without acquiring the information the doctrine itself required, and that he did not know he was missing it.

The law is written with blanks in it

This is not an accident of a few cases. It is a structural feature of how Anglo-American law is drafted, and it goes back much further than intellectual property.

Consider what the operative terms in most commercial and IP doctrine actually are. Likelihood of confusion. Secondary meaning. Fame. Substantial similarity. The ordinary observer. Transformative use. Custom and usage of the trade. Ordinary course of business. Material adverse effect. Reasonable expectations. Not one of these is defined in a way that permits application by reading. Each of them is a variable that a decisionmaker fills in with a set of facts about how people behave in some particular corner of the world.

Common law systems have always done this deliberately. When Lord Mansfield set about incorporating the law merchant into English common law in the second half of the eighteenth century, his method was not to reason from first principles about what commercial rules ought to be. He empaneled special juries drawn from the London merchant community and asked them what the practice was. He had a group of merchants who sat repeatedly and became, in effect, a standing advisory body on commercial custom. Mansfield understood something that gets lost in modern legal education, which is that the law of contracts is largely a machinery for enforcing expectations that the law itself does not generate. The expectations come from the trade. The court’s job is to find out what they are.

The Uniform Commercial Code made this explicit in the twentieth century, instructing courts to read agreements in light of course of performance, course of dealing, and usage of trade, and providing that these may supplement or qualify the express terms. Karl Llewellyn, the Code’s principal architect, was a legal realist who thought that pretending contract language had meaning independent of the commercial context in which it was used was a species of self-deception.

So when a lawyer says that cultural fluency matters, the claim is not that it would be nice to be well-rounded. The claim is that the doctrine has holes in it, and that the holes can only be filled with information about how people actually live, and that a lawyer who does not have that information is not in possession of the facts the legal test calls for.

Trademark, which is entirely a doctrine about perception

Trademark is the cleanest example, because a trademark has no existence apart from what consumers think.

Take Coca-Cola Co. v. Koke Co. of America, decided by the Supreme Court in 1920. The defendant’s argument was that Coca-Cola had come to court with unclean hands, because the name was deceptive: the product no longer contained a meaningful quantity of coca, and the public was being misled by a word that made a chemical claim the beverage could not support. Holmes brushed this aside on grounds that were entirely about culture rather than chemistry. The name, he wrote, had come to characterize a beverage that people could buy at almost any soda fountain, and it now signified a single thing coming from a single source, well known to the community. What the word had once described was legally irrelevant, because it was no longer what the word meant to anyone who used it. The etymology had been overtaken by the practice.

A year later, Learned Hand decided Bayer v. United Drug, which held that “aspirin” had become generic and could no longer be enforced as a mark. The reasoning is worth reading closely, because it is a piece of applied sociology. Hand did not ask whether the word had become generic in some abstract sense. He divided the market into distinct consuming populations and asked what the word meant to each. To manufacturing chemists and physicians, “aspirin” still signified Bayer’s product, because those groups had encountered it as a proprietary name. To the general public, who had bought the tablets across a counter for years without ever being told the chemical name, “aspirin” simply was the name of the drug, and had never been anything else. Hand’s disposition tracked that split precisely, and he was candid that the outcome turned on how a particular class of ordinary people had come to use a word in ordinary life.

The doctrine of dilution has similar origins. It was essentially invented in a 1927 law review article by Frank Schechter, who was then counsel to the BVD Company, and who was arguing from an observation about modern advertising rather than from precedent. Schechter’s insight was that the value of a mark had ceased to be its capacity to identify a source and had become its selling power, the psychological hold it exercised over the buying public, and that this hold could be eroded by uses on unrelated goods that created no confusion at all. Massachusetts enacted the first anti-dilution statute two decades later; federal law caught up in 1995. A doctrine that now sits in the Lanham Act began as a lawyer noticing something true about the culture that the existing legal vocabulary had no way to express.

The modern cases have the same shape. When the Second Circuit took up Christian Louboutin’s red sole in 2012, the question it had to answer was whether a color applied to a particular part of a shoe had come to mean something to the people who buy expensive shoes, and whether that meaning survived when the whole shoe was red. You cannot answer that question by reading the registration. You answer it by knowing something about how fashion signals work, how a house builds recognition, and what a monochrome shoe communicates that a contrasting sole does not.

Jack Daniel’s Products v. VIP Products is the same problem in a more entertaining costume. The Court’s holding, which was unanimous, was narrow: the Rogers threshold does not apply when the accused party is using the mark as a designation of source for its own goods, whatever expressive content those goods may also carry. But the underlying question the litigation put to the courts was one that no statute could answer, which is when a joke about a brand stops being commentary and starts being branding. That is a question about how consumers hold two ideas at once. Everyone who sees a Bad Spaniels squeaky toy understands it is not whiskey. The doctrinal question is whether they also understand it to have come from somewhere, and whether the somewhere they have in mind is the distillery. Justice Kagan’s opinion reproduced photographs of the toy, which is itself an admission that the analysis could not be conducted in words.

Then there is Hermès v. Rothschild, which a jury decided in February 2023, awarding roughly one hundred and thirty three thousand dollars and finding that the MetaBirkins NFTs infringed and diluted the Birkin marks. The appeal was argued before the Second Circuit in October 2024 and remains, as of this writing, undecided, which tells you something about how hard the panel has found it. Reading the record, the most interesting fight is not about the Rogers test at all. It is the fight over expert testimony, in which the district court excluded the defendant’s art critic on the ground that his methodology was unreliable, while admitting the plaintiff’s economist to explain that the works sat within a recognizable submarket of digital brand NFTs. One side was told that its account of artistic convention was not evidence. The other side’s account of market convention was. Whatever one thinks of that ruling, it is a ruling about which kind of cultural expertise a court is prepared to hear, which is a different question from what the law requires.

Copyright, where the scope of a bargain is a question of custom

Copyright presents the same structure in two places, and both are less obvious than the trademark version.

The first is license interpretation. When a grant is written in one technological era and a new medium appears, the question of whether the old grant covers the new use is, in practice, a question about what the parties in that industry understood themselves to be trading. Judge Friendly’s opinion in Bartsch v. Metro-Goldwyn-Mayer in 1968 held that a 1930 grant of motion picture rights carried the right to license the film for television, and reasoned from what a sophisticated grantor in the 1930s would have expected. Three decades later, a district court reached the opposite result in Random House v. Rosetta Books, holding that a right to publish a work “in book form” did not reach ebooks, because the trade understood a book to be a physical object and had separately bargained for electronic rights in later agreements. The two cases are not inconsistent so much as they are two applications of the same method to two different industries at two different moments. In both, the dispositive evidence was what people in the business had assumed.

The second is fair use, and specifically the first factor. Andy Warhol Foundation v. Goldsmith is worth revisiting not for its holding but for the argument between the majority and the dissent, because the two opinions display different theories of what cultural knowledge is for. Justice Sotomayor’s majority narrowed the inquiry to the specific challenged use, which was the licensing of the Orange Prince image to Condé Nast for a magazine article about Prince, and found that this use shared the same commercial purpose as the licensing market for Lynn Goldsmith’s original photograph. Justice Kagan’s dissent, joined by the Chief Justice, argued that the majority had failed to understand what Warhol did, and made the case at length, with reference to Titian and Manet and the history of artistic borrowing, that the new work said something the original did not.

It is tempting for anyone who cares about art to side with the dissent, and Kagan’s opinion is the more pleasurable to read. But the majority’s move is also a piece of cultural reasoning, just aimed at a different culture. Sotomayor was reasoning about the market for editorial photography, about how photographers actually earn a living, and about what happens to that market if a downstream artist can license the same subject to the same magazine for the same purpose. Both opinions are arguments about how a world works. They simply disagree about which world is the relevant one, and that disagreement, rather than any question of statutory construction, is what the case is about.

Music infringement litigation makes the stakes tangible. The “Blurred Lines” verdict in 2015, affirmed on appeal in 2018, awarded the Gaye family millions on a theory that a jury could find infringement based on a constellation of unprotectable elements that together produced a similar feel. Musicologists were close to unanimous that the decision confused genre with composition, and the effect on the industry was immediate and chilling. Eight years later, Ed Sheeran defended a claim that “Thinking Out Loud” infringed “Let’s Get It On” by picking up a guitar in a Manhattan courtroom and showing the jury how many songs are built from the same four chords in the same order. He won, and the Second Circuit affirmed the related ruling in 2024. The difference between the two outcomes was not a change in the law. It was that in the second case, the factfinder was given the cultural context necessary to distinguish a building block from a composition.

The right of publicity, which is a doctrine about recognition

The right of publicity has always been the most nakedly cultural of the IP regimes, because its subject matter is the fact of being recognized.

Its American origin is a failure. In 1902, the New York Court of Appeals held in Roberson v. Rochester Folding Box Co. that a young woman whose photograph had been reproduced on thousands of flour advertisements without her consent had no cause of action, because no such right existed at common law. The public reaction was severe enough that the legislature enacted a privacy statute the following year, which remains the basis of New York publicity law. In 1953, Judge Jerome Frank, deciding a dispute between chewing gum companies over the exclusive right to put baseball players’ photographs on cards, coined the term “right of publicity” and observed that prominent people would not be satisfied with mere damages for hurt feelings, because what they wanted was the money.

The doctrine’s expansion since then has been driven entirely by courts working out how recognition actually operates. Bette Midler recovered against Ford in 1988 not because anyone used her name or her image, but because an advertising agency, having failed to license her, hired one of her former backup singers to imitate her, and the Ninth Circuit understood that for a distinctive singer, the voice is the identity. Tom Waits recovered on similar grounds four years later. In White v. Samsung, a majority found that a robot in a blond wig turning letters on a game show set could appropriate Vanna White’s identity, prompting Judge Kozinski’s dissent from denial of rehearing en banc, which remains the best short essay any American judge has written about the costs of overprotecting cultural material. That entire line of cases exists because judges were willing to reason about how audiences recognize a person, rather than confining themselves to the enumerated attributes in a statute.

Which brings us to the present. In April 2023, an anonymous producer released a track called “Heart on My Sleeve” using AI-generated approximations of Drake’s and the Weeknd’s voices, and it accumulated millions of plays before it was removed. It was not clear that any existing federal cause of action fit cleanly. Tennessee responded in March 2024 with the ELVIS Act, which extended the state’s publicity statute to cover voice explicitly and to reach the tools used to produce imitations. Federal legislation remains pending. Anyone advising in this area right now is operating in the same position as the judges in the Midler line, which is to say they are reasoning from how audiences actually experience identity, in advance of a statute that describes it.

Contract, where this has been true the longest

None of this is peculiar to intellectual property, and it is worth making the point because it is the part most likely to be dismissed as soft.

In 1960, Judge Friendly, sitting by designation in the Southern District, decided a dispute between a Swiss buyer and an American seller over two shipments of poultry. The buyer had expected young broiling chickens and received older stewing fowl. Friendly opened by identifying the question as what the word “chicken” meant, and then spent the opinion examining trade usage, Department of Agriculture definitions that the contract had incorporated by reference, the negotiating correspondence, and, decisively, the price. The contract price was below the market price for broilers of that weight at the time, which meant that on the buyer’s reading, the seller had agreed to sell at a loss. Friendly concluded that the buyer had failed to carry its burden of proving the narrower meaning. The case is taught as a lesson in ambiguity. It is better read as a demonstration that the meaning of a commercial term is an empirical fact about a trade, discoverable by evidence, and that the most probative evidence was a number that only made sense to someone who knew what chickens cost.

Delaware’s merger litigation runs on the same fuel. When a buyer claims a material adverse effect, the court is asked whether a change has substantially threatened earnings potential in a durationally significant way, which requires the judge to form a view about what is normal for the industry over what horizon. When the pandemic produced a wave of cases about whether operational changes breached ordinary course covenants, the Court of Chancery had to decide what the ordinary course of business is for a hotel company in a period when nothing was ordinary, and it did so by looking at what comparable operators actually did. These are not exercises in reading. They are exercises in knowing an industry well enough to say what its baseline is.

Why this is getting harder

Three things have changed, and each of them raises the value of fluency rather than lowering it.

The first is fragmentation. Trademark doctrine developed in an era with a mass audience, when it was coherent to ask what the reasonably prudent consumer would think, because there was a broadly shared set of cultural references. There is no longer a general public in that sense. There are overlapping communities with their own vocabularies, and a term can be famous within a subculture of two million people and completely unknown outside it. Whether a mark is famous for dilution purposes, which the statute frames in terms of the general consuming public, is now a question that the statute’s own framing makes difficult to answer honestly.

The second is speed. The interval between a cultural practice emerging and a dispute about it reaching a court has collapsed. When Grand Upright was decided, sampling had been an established practice for over a decade, and the judge still did not know what it was. A dispute over a form that is eighteen months old gives everyone involved considerably less time to become informed.

The third is that a growing share of clients are not operating within an industry custom at all. They are building the category. For a client of this kind, industry convention is a data point about how incumbents behave, not a standard of care and not a guide to what their deal should look like. The lawyer’s job is to understand the emerging practice well enough to draft for it, which requires knowing what the client is actually building rather than mapping it onto the closest familiar thing. Most of the serious errors I have seen in this work come from that mapping reflex, which is a form of pattern-matching that feels like expertise and functions as its opposite.

Why the profession is structurally bad at this

The obvious explanation is that lawyers are busy, and it is true as far as it goes, but it is also the least interesting version of the argument and the one most likely to put a reader on the defensive. The better question is why a profession full of curious and intelligent people has ended up systematically underinvested in a form of knowledge that its own doctrines require.

Part of the answer is that the billable hour has no code for it. An afternoon spent reading about a music scene, an evening at a gallery opening, two hours following an argument playing out among the people who make the thing your client makes: none of this can be attributed to a matter, none of it can be captured, and in the internal accounting of most firms it is therefore indistinguishable from not working. Over the course of a career that produces a real selection effect, because the activities that build cultural knowledge are precisely the activities that a system measuring output as recorded time will treat as leisure.

Part of it is training. The case method teaches you to read an opinion by extracting the holding and discarding the facts, on the understanding that the facts are disposable packaging around the rule. That is a defensible way to learn doctrine and a poor way to practice in areas where the facts are the operative content of the rule. A lawyer trained to strip out the particulars will read Bayer v. United Drug as a case about genericness, when it is more usefully read as a case about two populations of people who had encountered the same word under different circumstances and therefore meant different things by it.

Part of it is how firms scale. Leverage depends on work that can be pushed down and templated, and particularized cultural knowledge is the least leverageable asset a firm can hold. It sits in one person’s head, it cannot be written into a form file, and it depreciates quickly enough that what you knew about a market in 2019 may now be actively misleading. Doctrinal knowledge compounds. This kind requires maintenance, which is unglamorous, unrecorded, and rewarded by nobody’s compensation model.

And part of it is that the profession has developed a workaround that conceals the problem rather than solving it, which is to buy the knowledge at trial. Retain a survey expert, a musicologist, an art historian, a market economist. This is often necessary and sometimes decisive, and it also arrives years after the point at which it would have mattered most. The judgments that determine the shape of a dispute are made long before anyone retains an expert, when someone decides whether to send the letter, which claims to plead, which forum to choose, how to characterize the client’s conduct, and what the client should have been doing differently two years earlier. Those calls are made by a lawyer reading a file, and the expert is hired afterward to support a theory that was selected without them. The Hermès record is a reminder that even at trial, a court may decline to hear the cultural expertise you have paid for while admitting your opponent’s.

The exposure is also uneven across practices. A firm whose clients are established companies operating inside settled industry custom will rarely notice the gap, because for that work the custom is knowable from documents and the relevant expectations have already been written down by someone. The gap opens when the client is doing something the custom does not yet describe, and it opens widest when the client’s entire enterprise depends on the difference between what they are making and what came before it.

None of this is a claim that a lawyer who does not follow music or design is a worse lawyer. A great deal of legal work has no cultural dimension whatsoever, and nobody drafting a credit agreement is handicapped by indifference to contemporary art. The claim is narrower and, I think, harder to argue with: for a particular band of work, this knowledge is a technical input to the analysis rather than an ornament around it, and the way the profession is organized makes it difficult to acquire and impossible to bill.

What fluency actually consists of

It is not a matter of liking the right music or knowing which designers matter this season, and the version of this argument that reduces to being culturally cool is a red herring!

It is closer to what an anthropologist does, which is to develop an accurate description of how a particular group of people behave and what their behavior means to them, and to keep that description current. In practice this means reading widely and outside the profession, paying attention to what clients are reading and watching and arguing about, spending time in the rooms where the work is made rather than only in the rooms where it is negotiated, and treating a client’s account of their own field as evidence to be taken seriously rather than as color to be translated into legal categories as quickly as possible.

It also means recognizing when you do not have it. The most useful thing about the Grand Upright opinion, read at a distance of three decades, is not that the judge was wrong about hip-hop. It is that nothing in the opinion suggests he considered the possibility that there was something to know. A lawyer who understands that the doctrine is full of blanks, and that the blanks are filled with facts about human behavior, will at least ask the question, and will know to go find someone who can answer it when the answer is not already in the file.

I want to ask a narrower question, one that the coverage has mostly skipped. Rebel Creamery sells sugar-free ice cream to people counting carbohydrates. Van Leeuwen sells butterfat and egg yolk to people who are not. The two companies sit in the same freezer case and almost nowhere else. Judge Eric Komitee found, after a bench trial, that shoppers were confused by the cartons, and I have no quarrel with that finding. But when he ordered Rebel to pay over $23.8 million of its profits, what exactly was changing hands, and on what theory?

To answer that you have to know something about how trademark law came to have a money remedy at all, which turns out to be a stranger story than the case summaries suggest.

The remedy came from equity, and it shows

Trademark protection in the United States grew out of the common law action for deceit and its equitable cousins, and for most of the nineteenth century the thing a mark owner actually wanted from a court was an order to stop. Money, when it came, came through the chancery side, as an accounting: the infringer was treated as having held its gains for the owner’s benefit, and was made to disclose and turn them over. That inheritance still governs. A claim for an infringer’s profits is restitutionary. It asks what the defendant got, not what the plaintiff lost, and because it sits in equity there is no Seventh Amendment right to have a jury decide it.

The Supreme Court gave the modern accounting its shape in 1916 in Hamilton-Brown Shoe Co. v. Wolf Brothers & Co., 240 U.S. 251, a fight between two shoe manufacturers over “The American Girl” and “American Lady.” The master’s report in that case reads like a piece of industrial archaeology: 974,016 pairs of shoes in one class, yielding $254,401.72 in profit, another 961,607 pairs in a second class, yielding $190,909.83. Hamilton-Brown argued that most of that money came from the shoes rather than the name, and that Wolf Brothers should have to prove which dollars were attributable to the mark. Justice Pitney rejected the premise. Where the defendant is not an innocent infringer, and where separating the profit owed to the mark from the profit owed to the merits of the goods is inherently impossible, the burden of that impossibility falls on the wrongdoer.

Frankfurter refined the point in 1942 in Mishawaka Rubber & Woolen Manufacturing Co. v. S.S. Kresge Co., 316 U.S. 203, a case about a red plug embedded in the center of a rubber heel. The opinion contains the sentence that every trademark brief has quoted since, about the owner’s right to redress when a competitor poaches the commercial magnetism of a symbol the owner created. What matters more for our purposes is the paragraph that follows, where Frankfurter concedes that awarding all of an infringer’s profits may hand the mark owner a windfall, and answers that the alternative hands the windfall to the infringer instead. That is not a theory of compensation. It is an allocation of uncertainty, and it was made on frankly moral grounds.

The Lanham Act absorbed all of this in 1946. Section 35, now 15 U.S.C. § 1117(a), lets a prevailing plaintiff recover the defendant’s profits, its own damages, and costs, subject to the principles of equity, and it codifies Hamilton-Brown’s allocation directly: the plaintiff proves the defendant’s sales, and the defendant must prove every element of cost or deduction it wants credited. The court may enter judgment for up to three times actual damages, or adjust a profits award up or down to a sum it finds just, provided the result compensates rather than penalizes.

The rest of the menu

Everything else in the statute arrived later, and each addition responded to a specific failure of the general rule.

Corrective advertising damages, which pay the senior user to buy back the public’s understanding of its own mark, come from Big O Tire Dealers, Inc. v. Goodyear Tire & Rubber Co., 561 F.2d 1365 (10th Cir. 1977). A Colorado tire cooperative had used “Big Foot” on two lines of tires. Goodyear, having learned of that use on the eve of a national campaign for which the money was already spent, went ahead anyway. The jury awarded $2.8 million in compensatory and $16.8 million in exemplary damages. The Tenth Circuit cut the compensatory figure to $678,302 by taking the share of Goodyear’s advertising spend that fell in Big O’s territory and then applying the Federal Trade Commission’s rule of thumb that undoing a false impression does not cost a dollar for every dollar spent creating it. Hold onto Big O. We will need it again at the end of this piece!

Statutory damages for counterfeiting came in 1996, in the Anticounterfeiting Consumer Protection Act, and now live at § 1117(c). They exist because counterfeiters do not keep books, and an accounting is worthless against a defendant with no records to account for. A plaintiff may elect a sum between $1,000 and $200,000 per counterfeit mark per type of goods, and up to $2 million where the use was willful. Anyone who runs a Schedule A enforcement program knows that this provision, and not the common law of confusion, is the engine of the whole practice. Section 1117(d), added with the Anticybersquatting Consumer Protection Act in 1999, does the same work for domain names, at $1,000 to $100,000 apiece.

Then there is willfulness, which for decades determined whether a plaintiff got to the accounting at all. The Second, Eighth, Ninth, Tenth and D.C. Circuits treated a finding of willful infringement as a precondition to any profits award. The Third, Fourth, Fifth, Sixth, Seventh and Eleventh did not. The Supreme Court resolved the split in 2020 in Romag Fasteners, Inc. v. Fossil Group, Inc., 140 S. Ct. 1492, where a jury had found that Fossil acted callously but not willfully, and the district court accordingly gave Romag nothing. Justice Gorsuch, writing for a unanimous Court, held that the statute nowhere makes willfulness an inflexible precondition, while adding that a defendant’s mental state remains a highly important consideration in deciding whether an award is appropriate. Romag opened a door and declined to say how wide.

Five years later the Court narrowed a different one. In Dewberry Group, Inc. v. Dewberry Engineers Inc., 604 U.S. ___ (2025), a district court had reached a $43 million profits award against a defendant that recorded no profits, by adding up the earnings of separately incorporated affiliates that nobody had bothered to name as defendants. Justice Kagan, again for a unanimous Court, vacated it. The defendant’s profits are the defendant’s profits.

Liability is common. Money is rare.

Set the statute aside and look at what actually happens to trademark plaintiffs who win.

Gucci sued Guess in 2009 over four marks and a trade dress, litigated for three years before Judge Scheindlin in the Southern District of New York, prevailed on most of what it pleaded, and asked for more than $221 million. It recovered $4.7 million. Scheindlin declined to find counterfeiting, on the ground that courts have confined that label to products copied stitch for stitch, and closed the opinion with a line from Oscar Wilde about fashion being a form of ugliness so intolerable that it must be altered every six months.

Hermès won its jury trial against Mason Rothschild over the MetaBirkins in January 2023, on infringement, dilution and cybersquatting, in what was treated at the time as a landmark for brands in digital markets. The jury awarded $110,000 in lost profits and $23,000 in statutory damages.

A&H Sportswear spent seven years litigating against Victoria’s Secret over MIRACLESUIT and THE MIRACLE BRA, traveling between the district court and the Third Circuit three separate times. It ended with an injunction and no monetary relief at all.

The reasons for this pattern exist. Proving that a particular sale was diverted by a particular design element is close to impossible when purchases turn on price, flavor, diet, shelf position and habit. Apportionment defeats most plaintiffs even when liability is clear. Courts remain reluctant to order a full accounting absent something that looks like bad faith, and Romag, whatever its text says, has not obviously changed that instinct.

Which is why the remedy that matters in most trademark cases is not in the damages section of the statute at all.

The injunction is the case

For the overwhelming majority of brand owners, winning means an order that the other side stop. That order is worth more than the judgment, and it is often worth more than the plaintiff could ever have proved in dollars. Rebel now has to redesign its packaging, revalidate it with retailers, reprint, replace point-of-sale material, rebuild shelf recognition, and absorb whatever share of its customers do not follow the new carton. Over several years that may cost more than the $23.785 million line item, and none of it appears in the coverage.

For about fifteen years the injunction was arguably at risk. In eBay Inc. v. MercExchange, L.L.C., 547 U.S. 388 (2006), the Supreme Court held that a patentee who wins is not automatically entitled to an injunction, and that courts must work through the traditional equitable factors case by case. Lower courts imported eBay into trademark. Mark Lemley’s 2017 article on the subject, Did eBay Irreparably Injure Trademark Law?, 92 Notre Dame L. Rev. 1793, made the argument that the transplant ignored what is different about trademark: injury to a brand is hard to quantify by its nature, damages awards are rare to begin with, and there is a third party in the room, the consumer, whose interest in not being confused is not represented by either litigant. Lemley’s sharpest observation is the one about compounding. Courts expanded eBay in trademark cases during the same period in which they were denying monetary relief, so that a mark owner could prove infringement and walk out with nothing whatsoever.

Congress fixed this in December 2020, in the Trademark Modernization Act, which amended § 1116(a) to give a plaintiff a rebuttable presumption of irreparable harm on a finding of violation for permanent relief, or on likelihood of success for preliminary relief. It is the least discussed and probably the most consequential trademark legislation of the last twenty years.

What happened in Brooklyn

Now the case.

Van Leeuwen adopted its pint design in 2016, after hiring Pentagram to prepare the brand for national wholesale. It claimed four elements in combination: monochromatic cardboard with a matching lid, a primarily pastel palette, black script lettering with an oversized first letter, and an overall minimalist design. Rebel argued that terms like “primarily pastel” and “minimalist” are too subjective to define enforceable rights, and separately that the individual components are ordinary throughout the category, which is true and which was beside the point. Komitee assessed the combination and the commercial impression it creates, found the claimed dress nonfunctional and inherently distinctive, and noted that competitors have many other ways to package premium ice cream. His articulation analysis is among the first detailed applications of the Second Circuit’s February 2025 decision in Cardinal Motors, Inc. v. H&H Sports Protection USA Inc., which held that articulating a trade dress with precision is a separate and lighter requirement than pleading its distinctiveness.

On confusion, a survey put net confusion at 34.3 percent, and there was evidence of confusion in the actual marketplace. Rebel’s argument that keto shoppers and premium dairy shoppers are different people, buying for different reasons, did not survive the fact that they stand in front of the same freezer door. It’s one of those arguments that sounds good in the office (because there’s legal support for it), but doesn’t quite work in the courtroom (because common sense prevails!).

What decided the case, though, was the evidentiary asymmetry. Van Leeuwen arrived with the entire Pentagram record: the brief, the research, the seven concepts presented, the ones rejected, the successive rounds. Rebel arrived with a finished Illustrator file and nothing behind it, no sketches, no earlier versions, no trail. Its founders testified that they had arrived at the design independently. Komitee found the probability of independently reaching that particular combination infinitesimal and called the account clearly fabricated. Barton Beebe’s empirical work on the multifactor tests, published in 2006, found that a finding of bad faith operates in practice as a nearly unrebuttable presumption of likely confusion, whatever the doctrine says about its being one factor among many. This case is a clean illustration.

How the number was built

Here is the part worth slowing down for.

Van Leeuwen never recovered a dollar of its own losses, and could not have. During discovery it declined to provide a computation of its damages under Rule 26(a), stating that such a computation was premature, and it never supplied one. Rule 37(c) then barred it from putting a damages calculation in front of the factfinder. Rebel moved to strike the jury demand on the ground that nothing legal remained in the case, and in March 2024 Komitee agreed. Under Gucci America, Inc. v. Weixing Li, 768 F.3d 122 (2d Cir. 2014), a Lanham Act accounting is equitable even when the plaintiff offers it as a stand-in for damages, and under Design Strategy, Inc. v. Davis, 469 F.3d 284 (2d Cir. 2006), a plaintiff who cannot prove actual loss has no jury right. So the case was tried to the bench, on an equitable remedy, because of a discovery choice made years earlier.

That choice matters for what the award means. The Second Circuit recognizes three distinct rationales for an accounting: avoiding unjust enrichment, standing in as a rough proxy for the plaintiff’s damages, and deterring infringement. See 4 Pillar Dynasty LLC v. New York & Co., 933 F.3d 202, 212 (2d Cir. 2019). Van Leeuwen expressly invoked the second. Its position throughout was that Rebel’s profits were a rough measure of Van Leeuwen’s own injury.

The arithmetic followed the burden allocation that Hamilton-Brown established in 1916. Van Leeuwen proved the revenue from the affected pints. Rebel then had to prove its deductions and, the harder task by far, to prove what portion of its profits was attributable to something other than the infringement. Rebel’s answer was that people bought its ice cream because it has no added sugar, not because the carton is lavender, which is an argument with real force. Komitee found that Rebel failed to carry its burden on apportionment, put the profit pool at roughly $35.5 million, and reduced it by a third to reflect demand attributable to the keto and better-for-you positioning. Under George Basch Co. v. Blue Coral, Inc., 968 F.2d 1532 (2d Cir. 1992), the defendant’s proven sales go to the plaintiff unless the defendant shows the infringement had no relationship to those earnings, so the residual uncertainty landed where the statute puts it.

I would watch that one-third on appeal because profits awards, though reviewed only for abuse of discretion, get remanded when an appellate court cannot follow the path to the figure.

The question nobody is asking

Which brings me back to where I started.

Komitee found a risk of reverse confusion, the doctrine born in Big O and adopted for the Lanham Act by the Second Circuit in Banff, Ltd. v. Federated Department Stores, Inc., 841 F.2d 486 (2d Cir. 1988). Reverse confusion describes the situation where a larger junior user saturates the market with something resembling a smaller senior user’s mark, so that the public comes to associate the mark with the newcomer and the originator starts to look like the imitator. In this case Rebel’s expansion into national retail threatened to leave Van Leeuwen, the company that commissioned the design in 2016, in the position of the copycat.

The doctrine’s whole premise is that the junior user is not trying to borrow anyone’s goodwill. It is building its own. Jeremy Sheff put this precisely in Reverse Confusion and the Justification of Trademark Protection, 30 Geo. Mason L. Rev. 123 (2022): the defendant in a reverse confusion case gains nothing the plaintiff would otherwise have enjoyed, no diverted sale and no reflected goodwill, and simply avoids the inconvenience of designing around. The harm to the senior user is real and can be severe, but it is a byproduct rather than a transfer. Sheff argues that this makes reverse confusion a poor fit for both of the standard justifications for trademark liability, the misappropriation account and the search-costs economics associated with Landes and Posner, since the junior user is not reaping where it has not sown and is, on any first-order welfare measure, lowering search costs for far more consumers than it confuses.

Now put that next to an accounting of profits, a remedy whose oldest and most secure justification is unjust enrichment. If Rebel’s profits came from keto shoppers buying a product Van Leeuwen does not make, then the enrichment did not come out of Van Leeuwen’s pocket, and disgorgement is doing something other than restitution.

The Seventh Circuit saw this thirty-four years ago and said so. In Sands, Taylor & Wood Co. v. Quaker Oats Co., 978 F.2d 947 (7th Cir. 1992), a Vermont company that owned THIRST-AID won against Quaker over the slogan “Gatorade is Thirst Aid,” and the district court awarded ten percent of Quaker’s Gatorade profits, $24,730,000, which with interest and fees came to more than $42 million. The Seventh Circuit affirmed liability and vacated the money. It observed that in a reverse confusion case the defendant is by definition not palming off, so the intent factor is essentially irrelevant, described the evidence of bad faith as pretty slim, and held the award inequitable, a windfall. On remand the court directed that a reasonable royalty serve as the baseline, with the defendant’s profits relevant chiefly to deterrence. See 34 F.3d 1340 (7th Cir. 1994).

The parallel is close enough to be uncomfortable. A reverse confusion finding, a disgorgement of the junior user’s profits, and a figure within a million dollars of the one Judge Komitee reached.

There are three answers available to Van Leeuwen, and they are not all equally good. The first is that this is not a pure reverse confusion case. Komitee found deliberate copying and fabricated testimony, which is forward-looking bad faith of exactly the kind Sands said was absent in Quaker, so the equitable objection has much less purchase. The second is that Van Leeuwen never asked for unjust enrichment in the first place. It sought profits as a proxy for its own damages, the second 4 Pillar Dynasty rationale, and Sheff’s critique goes to the first. The third is the oldest and probably the strongest: Frankfurter already told us, in 1942, that when the profit cannot be divided the windfall belongs to the owner rather than the wrongdoer, and Rebel had the burden and did not carry it.

Still, notice what those three answers have in common. Each of them is a reason the award survives despite the reverse confusion finding rather than because of it. The finding did important work on liability, where it answered Rebel’s best defense, that its customers and Van Leeuwen’s are different people who want different things. On remedy it does no work at all, and arguably cuts the other way. That gap between the theory of the injury and the theory of the payment is not a flaw in Judge Komitee’s opinion. It is a flaw in the law he was applying, and it has been sitting there, unexamined, since 1977.

Whether the Second Circuit gets a chance to look at it depends on what Rebel does next.

---

*Sources: Van Leeuwen Ice Cream LLC v. Rebel Creamery LLC, No. 1:21-cv-02356 (E.D.N.Y.), Dkt. 65 (Mar. 11, 2024) and Dkt. 125 (July 16, 2026). Case citations as noted throughout.*

The thing Phia is accused of doing takes about two hundred milliseconds, and you would never see it.

You are at checkout, buying a coat. You found the coat because a woman whose taste you admire wore it in a photograph, and you clicked her link, and a small marker was left on your computer recording that she was the reason you were there. That marker is how she gets paid. It is, in a fairly literal sense, her entire income. According to a Bloomberg investigation published on July 9, the Phia browser extension, at the moment you were reaching for your credit card, opens a tab you never saw, inserts its own referral code where hers had been, and takes the commission. The behavior was allegedly observed across more than fifty online retailers.

Phia is a shopping extension founded in 2025 by Phoebe Gates and Sophia Kianni, who met as roommates at Stanford. It compares prices across new and resale inventory and hunts for discount codes, and it is aimed, in Gates’s own framing, at the young woman who shops like a genius and does not want to waste time doing it. It has raised more than forty three million dollars at a valuation around one hundred eighty five million, with backing that includes Khloé Kardashian and Hailey Bieber. Gates is twenty three, and she is the youngest daughter of Bill Gates and Melinda French Gates, a fact she has spent two years trying to make irrelevant. She has taken no money from her parents. She has said she wants Phia to succeed with no ties to her privilege or her last name.

Impact.com, one of the largest affiliate platforms, suspended Phia’s account for conduct inconsistent with its policies and began reviewing the affected transactions. Phia says the problem has been fixed, and Bloomberg’s follow up testing found that it had been.

That is the news, and the news is the least interesting part.

What makes the Phia story worth the attention of anyone who works at the intersection of fashion and law is that the trick it is accused of is not novel, not clever, and not even particularly well hidden. It is thirty years old. A man went to federal prison for it in 2014. PayPal is in discovery over it right now. The technique persists because the thing it exploits is not a bug in anyone’s code. It is the foundational accounting fiction of the entire creator economy, and the online fashion business, which depends on that fiction more completely than any other industry on earth, has never once looked at it directly.

The ledger

Somewhere in the files of the United States Patent and Trademark Office sits patent number 6,141,666, issued on Halloween of 2000 to a man named William J. Tobin. Tobin was an Army veteran who had begun his working life mowing lawns in New York with an edger his mother bought him, and who went on to found ten companies. The seventh sold flowers. In 1989, on the Prodigy network, when getting online meant listening to a modem negotiate with the void, he built an arrangement in which he paid Prodigy a cut of every sale Prodigy sent him. By 1993 it was generating six million dollars a year. Without fanfare, and years before anyone thought to name it, Tobin had invented the business model that now underwrites most of what you read on the internet for free.

The idea is old as commerce. Send me a customer and I will pay you a slice of what they spend. The maître d’ takes a drink for seating you at the good table. What was new was that it could be counted.

The counting is where Phia lives.

In 1994 an engineer at Netscape named Lou Montulli devised a small packet of data that a website could leave on your machine, a marker that would survive after you closed the window and would tell the site, on your return, who you were and where you had been. He called it a cookie. He was solving a mundane problem, the shopping cart that forgot its contents. What he actually built, almost by accident, was the ledger of the internet’s entire economy. Every claim any publisher has ever made to a commission rests on that marker and on the assumption that it tells the truth.

CDNow launched BuyWeb the same year, paying music sites that linked to albums they had reviewed. Amazon followed in 1996 with Associates, not the first program of its kind but the one that made the model legible to everyone else. Commission Junction and ClickBank arrived in 1998 to sit in the middle and verify the count. Within a decade of Tobin’s roses, an invisible payments layer had been laid beneath the visible web, and it ran on cookies. It still does. Phia’s alleged offense is a crime against a ledger that a Netscape engineer invented to remember your shopping cart.

Why fashion has the most to lose

For a long time fashion regarded all of this as beneath it, which is among the more instructive facts in the history of retail.

In 2010 a young woman in Dallas named Amber Venz Box was working as a personal shopper and writing a fashion blog. She had done the usual apprenticeships: fit model, intern at Thakoon, assistant buyer at a luxury boutique. She had also made a mistake she did not understand until the Dallas Morning News wrote about her and sent readers to her site. Her clients read the blog, absorbed her taste, and went off to buy the clothes themselves. In the boutique she had been paid a commission for exactly this work. Online she was doing the same labor for nothing.

That is the primal scene of the creator economy, and notice what it is actually about, which is whether taste is work.

Venz Box and her then boyfriend Baxter Box built the tracking links themselves and launched rewardStyle in 2011 with a few hundred dollars in the bank. Engineering was never the hard part. The retailers were, and they told her, in the tone the industry reserves for people it considers unserious, that they worked with celebrities and models, that they were elevated, that a woman with a website was not part of the plan. ShopBop and Net a Porter said yes first. The rest followed, because numbers do what numbers do.

What that company became, after LIKEtoKNOW.it and the rebrand to LTK and three hundred million dollars from SoftBank at a two billion dollar valuation, is now the circulatory system of consumer fashion. Something like six billion dollars a year in brand sales run through it. Three hundred thousand creators depend on it. The commissions, generally ten to twenty five percent, are the actual income of a very large number of people whose professional existence rests on the proposition that when a follower sees a coat on them and buys that coat, the sale belongs to them in some meaningful moral sense.

Fashion turned out to be the ideal substrate for affiliate marketing, for a reason that is about more than click rates. In most categories the affiliate is a middleman, a coupon site or a comparison engine adding convenience to a decision the shopper had mostly made already. In fashion the affiliate is frequently the entire reason the desire exists. Nobody wakes up needing a particular pair of loafers. Someone shows you the loafers, on their body, in their apartment, in a light you find enviable, and the want is manufactured on the spot. The recommendation is not adjacent to the product. The recommendation is the product.

Which means the cookie, in fashion, is arguably not a technicality. It is often the instrument by which persuasion gets paid.

And it means that the woman Phia is allegedly stepping in front of is not a coupon aggregator or a price bot. She is the reason the sale happened. Phia’s stated customer, the young woman who shops like a genius, is precisely the customer that LTK’s creators built. The extension arrives at the end of a chain of influence it did nothing to create and, if the allegations hold, signs its own name at the bottom.

Cookie Stuffing

Here is the flaw, and it has been in plain sight since Tobin’s roses.

The industry settled, for reasons of administrative convenience, on a rule called last click attribution. Whoever dropped the most recent cookie before purchase gets the money. Everyone knows it is a crude rule. A shopper might read a review in October, see three creators wear the thing in November, search for it by name in December, and buy it after passing through some coupon site on the final afternoon. Last click hands the whole commission to the coupon site and nothing to the four people who did the persuading. The industry adopted it anyway, because a crude rule that yields a definite answer is easier to administer than a subtle rule that yields ambiguity and an argument.

But a payment system that rewards whoever stands closest to the register at the moment of sale will attract people whose only strategy is to stand very close to the register. If the position can be simulated, it will be.

The technique is called cookie stuffing. You drop your affiliate cookie onto a browser that never clicked anything of yours, and you wait. Scatter enough cookies across enough machines and some of those machines will wander into the merchant on their own, and the merchant will consult the ledger, see your name, and pay you for a sale you had nothing to do with. It is not theft in the sense of taking something from a vault. It is a forgery in the register.

The seminal case is a strange one. In 2006 eBay noticed that its single largest affiliate was collecting extraordinary sums while sending eBay nothing of value. His name was Shawn Hogan, and he was not a figure from the internet’s demimonde. He ran a legitimate software company and hosted one of the largest webmaster forums of the era. He was, in the vernacular of the moment, a respected guy. What he had done was rig his sites to load a resource from eBay’s servers, setting an eBay affiliate cookie on every visitor with no click required. He limited it to once per user so the traffic would look organic. Over several years eBay paid him more than twenty eight million dollars.

The FBI got involved. A grand jury indicted Hogan in 2010 on ten counts of wire fraud. His counterpart, eBay’s second most productive affiliate, a man named Brian Dunning, had run a similar operation and taken millions of his own. Both insisted eBay had known all along and encouraged it. Neither could prove it. Hogan pleaded to a single count in December 2012 and was sentenced in May 2014 to five months in federal prison and a twenty five thousand dollar fine, which struck the industry as comically light against the sums involved. Dunning got fifteen months.

For lawyers, the interesting thing is that the wire fraud predicate was not the commission payments moving between states. It was the transmission of the cookies themselves, a few bytes crossing state lines and international borders on the way to a stranger’s browser. The government treated the forged marker as the fraudulent wire. That is elegant construction, and it converted what everyone had understood as a breach of a terms of service agreement into a federal felony. It also established the doctrinal fact that governs the Phia question entirely: a cookie is a statement, a false cookie is a false statement, and false statements transmitted across wires for money are a crime.

One more detail to carry forward. When Hogan and Dunning denied the stuffing to Commission Junction, prosecutors used the denials as evidence of intent. The cover up furnished the mens rea. It usually does.

The extension era

For a decade after Hogan, the industry treated cookie stuffing as solved, a species of fraud practiced by marginal actors and largely driven out by better detection. Researchers who went looking for it in the wild found little.

Then, on the weekend before Christmas in 2024, a YouTuber called MegaLag published a video about a browser extension owned by PayPal.

Honey was not a marginal actor. PayPal had bought it in January 2020 for roughly four billion dollars, drawn by seventeen million monthly users. It presented itself as a helpful little assistant that hunted for coupon codes at checkout, and it had been promoted, with enthusiasm and considerable payment, by exactly the population of creators whose livelihood depends on attribution.

The allegation was that when a shopper clicked Honey’s button at checkout, Honey opened a background tab functioning as a simulated referral, stuffed its own cookie, displaced whatever cookie the creator had earned, and collected the commission under last click. In the most baroque version of the accusation, Honey was doing this to the very creators it was paying to advertise Honey.

Anyone reading the Bloomberg reporting on Phia will recognize the architecture immediately. Background tab. Simulated click. Substituted code. The two products are accused of the same maneuver, executed in the same two hundred milliseconds, for the same reason.

The Honey litigation is the map of where Phia’s problem goes next, and it is not a simple map. In November 2025 a federal court denied PayPal’s motion to compel arbitration, and then dismissed the operative complaint with leave to amend on standing grounds, which is a lawyerly way of saying the plaintiffs had suffered a moral injury they had not yet learned to plead as a legal one. In January 2026 they returned with a second amended consolidated complaint attaching the actual merchant contracts, which is what you do when a court informs you that outrage is not a cause of action. The case is in discovery. PayPal denies wrongdoing and maintains that its attribution practices conform to industry standards, a defense that is either a complete answer or a devastating indictment of the industry, depending on how you hear it.

What Phia is actually accused of, and why the answer is not obvious

The Phia case is not so straightforward.

Three parties are plausibly injured. The creator who earned the referral and lost the commission has the most sympathetic claim and, as the Honey plaintiffs discovered last November, the hardest one to plead, because proving that this particular sale would have been credited to you requires reconstructing a counterfactual inside a stranger’s browser. The merchant is injured in a way that gets less attention: it paid a commission for a customer it already had, which means it bought nothing at all. And the affiliate networks suffer the injury that matters most in the long run, which is that their numbers stop meaning anything. That is presumably why Impact.com moved as fast as it did.

The theories available against Phia are the ones that have always been available. Breach of the affiliate agreement, which is straightforward and therefore cheap. Unfair competition and tortious interference, which reach the harm to the displaced creator. And, if intent can be shown, the wire fraud theory that Hogan’s prosecutors built, which remains the sharpest instrument in the drawer precisely because it treats the false cookie as the false statement.

Which brings us to the most consequential two words in the entire episode, and the ones a fashion lawyer should sit with: technical issue.

Phia’s position is that this was a bug, promptly fixed. That is not merely a public relations posture. It is a claim about mental state, and mental state is the whole ballgame in criminal law. Cookie stuffing done knowingly, for money, is fraud, and Shawn Hogan went to prison for it. Cookie stuffing produced by background tab logic written carelessly by a young engineering team in a hurry is a contract breach and an embarrassment (civil, not criminal). The distance between those two characterizations is the distance between a compliance headache and an indictment, and if it is ever litigated it will be litigated in the commit history, in the Slack messages, in the internal dashboards that either did or did not show a suspicious rise in attributed conversions that nobody thought to question. Hogan’s denials supplied his intent. Phia’s engineers have left a record of their own, and in due course it will say whatever it says.

There is a further wrinkle that makes this genuinely hard rather than merely scandalous. Last click attribution is a rule the industry wrote for its own convenience, and extensions occupy an ambiguous position under it. An extension that legitimately finds you a coupon has, in a thin sense, participated in the sale. PayPal’s defense amounts to the observation that everyone in this business has been standing near the register for years and nobody minded until a YouTuber filmed it, and that defense is not frivolous.

It is also not exculpatory, and the reason is the cleanest sentence anyone has produced in this whole affair. It comes from Ben Edelman, the affiliate researcher who helped test the Phia extension, and it deserves to be carved above the door of the industry: only a real click will do.

The fiction at the bottom

I keep returning to Amber Venz Box in 2010, watching her readers take her taste and spend money somewhere she could not follow.

What she built to solve that problem was, in the end, a legal fiction. The cookie does not record who persuaded you. It records who touched you last. The industry chose that proxy because it needed some rule, and a rule that can be computed beats a rule that must be argued. For fifteen years the fiction held, and a profession grew up inside it, and hundreds of thousands of people learned to make a living by being persuasive on the internet and trusting that a small text file would tell the truth about it afterward.

Cookie stuffing is what happens when someone notices that the fiction is a fiction. It is not a hack of the system. It is a rational reading of the system’s actual rules by people who declined to pretend along with everyone else. Hogan understood it in 2006. Honey, allegedly, understood it in 2020. Whoever wrote the offending lines at Phia understood it, or should have, in 2025.

Fashion has more riding on this than any other industry and is the least equipped to see it coming. Its entire creator economy sits on an attribution model that was never designed to measure influence and has always been trivially easy to counterfeit at the final moment. Every commission a stylist earns, every payout that lets a woman in Ohio employ fifteen other mothers to run her linking business, depends on a marker that a well capitalized extension can overwrite in the fraction of a second before checkout resolves.

So the remarkable thing about Phia is not that a young founder with a famous surname is accused of cutting a corner, and it is certainly not that a Gates is involved, which is the only part of this story most people will remember. The remarkable thing is that the corner has been in the same place, unfixed and unguarded, for thirty years, and that we have arranged the incentives so that walking past it without cutting it requires a restraint the industry has never once bothered to reward.

Phia may well have shipped a bug. That is the charitable reading and it may be the true one. But a system this easy to rob will keep getting robbed, by the careless and the deliberate alike, and it will not be possible to tell them apart from the outside.

Somebody is eventually going to have to fix the ledger. Until then we should stop being surprised at what keeps getting written into it.

In the winter of 1976, a jury in Denver did something that still unsettles the largest companies in America. A regional tire outfit called Big O had put the words BIG FOOT on the sidewall of a couple of its tire lines. Goodyear, the biggest tire maker on earth, later rolled out its own BIGFOOT campaign, spent something close to ten million dollars saturating the airwaves, and kept going, even after Big O told it to stop. Big O sued. The theory was strange enough that only one court had ever entertained it before: not that the little company was riding the giant’s coattails, but the reverse, that the giant’s advertising was so vast it would swallow the small brand’s identity whole, so that customers would assume Big O had copied Goodyear rather than the other way around. The jury sided with the small company and handed down an award that ran into the millions. The Tenth Circuit affirmed the liability, trimmed the number, and in the process gave the doctrine its name. Reverse confusion. A jury of ordinary people, asked what ordinary people would think, and trusted to answer.

Keep that scene in mind. Fifty years later, another small brand, this time a coffee company, would make more or less the same argument against another giant. And it would never see a jury at all.

* * *

The question underneath both cases is older than either, and it sounds, at first, like something you would argue about in a seminar rather than a courtroom. Is a given issue a question of fact or a question of law? Fact questions go to the jury, or to a judge sitting in the jury’s place, and they are reviewed on appeal with a great deal of deference. Law questions belong to the judge, and appellate courts review them fresh, owing the trial court nothing. The line between the two is one of the oldest fault lines in the Anglo American legal tradition, running back to the split between the courts of law and the courts of equity, where some claims arrived with a jury and some arrived only with a chancellor. Trademark grew up on both sides of that line. It was a creature of equity when a plaintiff wanted an injunction and a creature of law when a plaintiff wanted damages, and it has never fully decided which parent it takes after.

For most of the twentieth century, courts did not agonize over this, because trademark infringement had a single organizing question that everyone agreed was messy and factual: likelihood of confusion. Would an ordinary buyer, moving through the aisles at ordinary speed, mistake one brand for another. In 1961, Judge Henry Friendly, writing for the Second Circuit in a dispute between Polaroid and a company called Polarad, laid out a list of factors for weighing it. Other circuits built their own lists. The Ninth Circuit had Sleekcraft, the Fifth had its digits of confusion, and so on, but the bones were the same everywhere. Strength of the plaintiff’s mark. Similarity of the marks. Proximity of the products. Evidence of actual confusion. Good faith. The lists were long and no single factor controlled, which is another way of saying the whole exercise looked and felt like fact finding.

The first factor on nearly every one of those lists, and often the heaviest, is the strength of the senior mark. In 1976, the same Judge Friendly gave that idea its enduring shape in a fight over the word SAFARI, sorting marks along a spectrum from generic at the weak end, through descriptive and suggestive, up to arbitrary and fanciful at the strong end. APPLE for computers sits near the top, because the word tells you nothing about the machine. A word that hints at the product without quite describing it, that asks the buyer to make a small imaginative leap, is suggestive, and lives in the muddy middle. Strength turned out to have two components that the cases do not always keep straight. There is inherent, or conceptual, strength, which is a mark’s position on that spectrum, a question about the word itself. And there is commercial, or acquired, strength, which is about how much recognition the mark has earned in the market through years of use and advertising. The second kind is plainly about consumers and plainly factual. The first kind is where the trouble hides.

Deciding whether RISE is suggestive or merely descriptive, whether it is strong or weak, is not quite the pure legal question that reading a statute is, and not quite the pure factual question that counting confused customers is. It is a hybrid. Patent law ran into the same difficulty with the meaning of patent claims, and in 1996, in Markman, the Supreme Court called claim construction a mongrel practice, part fact and part law, before handing it to judges anyway. Trademark strength is a mongrel of the same litter. For decades the circuits treated its placement on the spectrum as a fact, sent it to juries, and reviewed it gently on appeal. One court did not.

* * *

The last decade or so is really a story about the Supreme Court making up its mind. Twice in ten years it took a trademark question that looked technical and answered it with the same instinct. In 2015, in Hana Financial, the Court considered a doctrine called tacking, which asks whether a business that tweaks its logo over the years can claim the earlier version’s priority date. It held, unanimously, that the question belongs to the jury, because it turns on the impression the marks leave on an ordinary consumer. In 2020, in the Booking.com case, the Court held that whether “Booking.com” is a generic name for an online travel service depends on what consumers actually understand the term to mean, which is to say a question of fact about perception, not a proposition you can deduce from the dictionary. The thread connecting the two is hard to miss. When the question is what the ordinary person thinks, the Court keeps giving it to ordinary people.

Against that current, the Second Circuit spent the same years swimming the other way. In a run of opinions, some of it in dicta and footnotes, it developed the view that the inherent strength of a mark, its place on Friendly’s spectrum, is a question of law for the judge. A stray line in a 2003 case grew into a footnote in a 2020 decision involving Tiffany and Costco, and hardened, case by case, into a rule. The reasoning was not unreasonable. Placing a word on a spectrum can feel like an act of legal classification, the kind of sorting judges do, and there is real value in letting them do it early, before the machinery of a full trial grinds into motion. But it left one of the most powerful courts in the country, the court that sits over Madison Avenue and most of the advertising industry, standing alone against every other circuit on a question the Supreme Court had twice leaned the other way.

The practical consequence played out on summary judgment, which is where modern trademark cases go to live or die. If the strength of the plaintiff’s mark is a fact for the jury, a judge can rarely resolve it on the papers, and the case grinds toward trial, with everything trial drags along behind it: discovery, consumer surveys that cost as much as a house, dueling experts, and years. If strength is a question of law, a judge can look at a suggestive word in a crowded field, pronounce it weak, and close the case in an afternoon. In an era when small brands have increasingly borrowed Big O’s old theory to sue giants for reverse confusion, that difference decides who ever gets to make their case to twelve human beings. I have sat through enough of these fights to tell you that the label we hang on the strength question is not a technicality. It is the whole shape of the litigation, and very often the whole leverage of the settlement.

* * *

Which brings us to the coffee. RiseandShine Corporation, which does business as Rise Brewing, sells nitro cold brew in a can under the mark RISE. In 2021, PepsiCo launched an energy drink called Mtn Dew Rise Energy, aimed, as the name suggests, at the morning. Rise Brewing sued, arguing that Pepsi’s enormous marketing muscle would flood the market and leave consumers believing the smaller company was the one trading on Pepsi’s name. It was Big O and Goodyear again, poured over ice. There was even a detail that would not be out of place in a novel: Pepsi, according to the reporting, had looked at buying Rise before it built its own version. The district court in Manhattan agreed the threat was serious and granted a preliminary injunction.

The Second Circuit vacated it. Applying its homegrown rule, the panel treated the inherent strength of RISE as a question of law and decided the word was weak on its face, because “rise” carries logical associations with coffee, with mornings, with the lift that caffeine promises, and because the shelves were already crowded with beverages using the term. On remand, the district court entered summary judgment for Pepsi, and in 2024 the Second Circuit affirmed. The mark had been declared weak as a matter of law. The confusion question, the one that in Denver had gone to a jury, never reached one here.

Rise Brewing asked the Supreme Court to take a single, clean question: whether the strength of a trademark is a question of fact in the likelihood of confusion analysis. Its argument leaned directly on Hana Financial and Booking.com. If tacking and genericness turn on the consumer’s mind, and therefore go to the jury, how can the strength of a mark, which turns on the same mind, belong to the judge alone? Twelve circuits treat a mark’s place on the spectrum as factual. One does not. That is a clean split, and the petition framed it cleanly.

The most telling move came from the government. The Court had invited the Solicitor General to weigh in, and the brief that came back did something you do not see every day. It agreed that the Second Circuit had gotten the law wrong, that inherent strength is not properly a pure question of law, and then urged the Court to decline the case anyway, on the theory that the error may not have changed the outcome, since other factors in the confusion analysis also cut against Rise. It was a lawyer’s brief in the best and most maddening sense: you are right, and you should still lose. In late June, as the term closed, the Court granted review regardless, over the government’s advice.

So what is actually on the table when the Justices hear this next term. On the surface, a standard of review and a procedural label. Whether an appellate court checks a strength ruling fresh or with deference. Whether these cases can end on summary judgment or must more often be tried. Underneath, something with more voltage. The Seventh Amendment preserves the right to a civil jury, and there is a live disagreement, including among the Court’s conservatives, between a reverence for that guarantee and a worry that juries need firmer judicial supervision. A case about cold brew and soda is going to become, in the marble room, a case about how much we trust ordinary people to say what ordinary people think.

For brands, none of this is abstract. If the Court keeps strength on the judge’s side of the line, expect more infringement suits to die early, more defendants willing to fight rather than settle, more shaky enforcement claims folding just before the motion is decided. Good news if you are defending a suggestive word in a crowded category. Hard news if you are the brand trying to protect one. If the Court moves strength to the jury’s side, the leverage inverts. More cases reach trial, plaintiffs can press claims that used to be dismissed on the papers, and defendants face longer, costlier fights with far less certainty about how any of it ends. Either way, the ruling will move the three questions every client actually asks, sometimes in these words and sometimes only with their eyes: how much, how long, and who decides.

The Denver jury that named this doctrine in 1976 went home a long time ago. What the Court decides next term is whether, in the one corner of the law that is entirely about the ordinary person’s mind, we still send the question to a room full of ordinary people, or whether that judgment now belongs to one person in a robe who has read the briefs and settled the matter before lunch. I know which answer I would give. I also know that some of the sharpest people in this field would give the other one, which is exactly why it is worth watching, and exactly why the can of coffee matters more than it looks.

The mark was the first technology of trust. Everything since has been an argument about how to defend it.

## The analog centuries

Medieval England took the problem seriously enough to legislate it. The Bakers Marking Law of 1266 required bakers to stamp their loaves so that a short-weighted loaf could be traced back to the baker who cheated, which is to say the earliest trademark law was really a consumer-protection law wearing a merchant’s coat. The cutlers of Sheffield registered their marks. The goldsmiths punched theirs. A mark was a promise, and a false mark was a lie told at scale.

The common law eventually caught up to the obvious. In *Southern v. How*, decided in 1618, an English court took note of an older dispute in which a clothier had sold inferior cloth under another maker’s mark, and the idea that this was actionable began its long migration into doctrine. By the nineteenth century the principle had hardened. *Sykes v. Sykes* in 1824 punished a maker who stamped a rival’s name on his shot belts. *Millington v. Fox* in 1838 went further and granted relief even where the copyist had not intended to deceive, which planted a seed that still grows in our law, the notion that consumer confusion, not bad intent, is the harm.

America imported the instinct and then federalized it, clumsily at first. Congress passed a trademark statute in 1870, and the Supreme Court struck it down in 1879 in *The Trademark Cases*, holding that Congress had hung the law on the wrong constitutional hook. The fix was the Commerce Clause, and the modern architecture arrived in 1946 with the Lanham Act, the statute under which nearly every fight in this essay is still waged.

For most of the century that followed, enforcement was a physical act. You hired investigators. You staged raids. You seized product off folding tables on Canal Street and out of containers at the port, and you measured a good year in pallets destroyed. Congress eventually made the worst of it criminal through the Trademark Counterfeiting Act of 1984, which put real prison exposure behind the trade. The system was expensive and local and slow, and it mostly worked, because the counterfeiter had to occupy physical space that you could find and a court could reach.

Then the supply chain dematerialized, and the table on Canal Street became a storefront with no address.

## The cases that built the internet’s rulebook

Three decisions did most of the work of deciding who pays when the fake moves online.

*Inwood Laboratories v. Ives*, in 1982, gave us the contributory infringement standard, the idea that someone who supplies a product to a party it knows is infringing can share the liability. It was written about drug manufacturers, but it became the lever for everything that followed.

*Tiffany v. eBay*, decided by the Second Circuit in 2010, then set the burden in a way brands have resented ever since. The court held that a marketplace is not liable for the fakes its users sell unless it has specific knowledge of specific listings. Generalized awareness that counterfeiting is rampant on your platform is not enough. The practical effect was to put the policing cost back on the brand, listing by listing, which is the origin of the whack-a-mole every general counsel now complains about.

*Gucci America v. Frontline Processing*, in the Southern District of New York the same year, pointed at the more interesting target. It held that a payment processor could be contributorily liable for knowingly handling a counterfeiter’s transactions. That is the choke-point theory, the recognition that you do not have to catch the seller if you can cut off the money, and it quietly reframed the next fifteen years of enforcement strategy.

## Schedule A, the engine of the modern era

Out of that landscape came the single most consequential procedural invention in trademark practice, and most people outside the field have never heard of it.

Sometime in the early 2010s, brands and their lawyers, working most often in the Northern District of Illinois, began filing a new kind of lawsuit. Instead of suing one seller, they sued hundreds at once, listing the anonymous online storefronts not in the complaint but in a separate, sealed exhibit titled Schedule A. They would then ask the court, ex parte and before any defendant knew a thing, for a temporary restraining order freezing the sellers’ accounts across Amazon, PayPal, Alipay and the rest, and seizing their domains.

The genius of it, and I use that word with mixed feelings, is that it attacks money and access rather than bodies. A defaulting Chinese merchant cannot be hauled into a Chicago courtroom, but his frozen PayPal balance can. Faced with a freeze and no capped amount, most sellers default or settle quickly just to thaw their accounts and get back to business. It is enforcement engineered for a world where the defendant is a username.

It is also, increasingly, contested. Professor Eric Goldman gave the practice its name, the Schedule A Defendants Scheme, or SAD Scheme, in a 2023 piece in the Columbia Law Review Forum, and he documented more than three thousand such cases and tens of millions of dollars in burden on the courts. The criticism is not frivolous. These cases routinely proceed without serving the complaint, without establishing personal jurisdiction, and against defendants joined together for no better reason than that they are all foreign and all online. In early 2025 a quartet of law professors, Goldman among them, urged the Federal Circuit to rein it in. Judges in the very district that birthed the tactic have begun to push back, severing misjoined defendants and looking harder at the ex parte freezes.

So the engine still runs, and the numbers show it. United States trademark suits jumped roughly twenty-five percent last year, to more than four thousand, a great deal of it this exact maneuver. But the smart money is already migrating, out of a skeptical Northern Illinois and toward friendlier venues, and tightening its filings to survive the scrutiny that is coming.

## What is actually working now

Step back from the courtroom and the picture changes, because the most effective anti-counterfeiting moves of the last few years did not happen in one.

Start with the demand side, because it is the part lawyers like to ignore. The fake is no longer shameful. The word dupe has become a flex, and the global trade in counterfeit goods now runs around four hundred and sixty-seven billion dollars a year by the OECD’s count, with projections reaching toward one and three-quarter trillion by the end of the decade. The buyers are not who you think. Something like seventy percent of American households earning over a hundred and fifty thousand dollars admit to buying dupes, outpacing every lower bracket. You cannot litigate your way out of a culture that has decided the knockoff is good taste.

What you can do is make the channel uneconomical, and that is where the real wins are.

The death of the de minimis exemption is the clearest example. For years, parcels valued under eight hundred dollars entered the United States with almost no inspection, and well over a billion of them did so in 2024 alone, which is the loophole that built the Shein and Temu logistics model and flooded the country with low-value fakes. When the government finally closed it in 2025, first for China and then globally, the effect was immediate in a way no injunction has ever been. Temu’s daily active users in the United States fell by more than half within weeks, and Shein’s dropped about a quarter. That is enforcement by customs schedule, not by complaint.

Platforms are the other front. The INFORM Consumers Act, in force since 2023, requires marketplaces to verify the identity of high-volume third-party sellers, which strips away some of the anonymity the whole counterfeit economy depends on. Amazon’s Brand Registry and its Counterfeit Crimes Unit, whatever you think of the company, have turned platform-level detection into something faster than any law firm’s takedown queue. The choke-point theory from *Frontline* has matured into a strategy, applied to payment processors and logistics providers alike.

## Where the fight goes next

The next few years belong to structural enforcement over adversarial enforcement. Choking logistics, pressuring platforms, and squeezing the payment rails will keep outperforming the one-defendant-at-a-time lawsuit, because they change the economics of the channel rather than punishing a single ant on it. Schedule A will survive, but disciplined, as courts force it to honor the procedural safeguards it has spent a decade routing around, and as filings migrate to venues that will still hear them.

The harder frontier is the one our law barely touches, and it is the one that matters most in fashion. A counterfeit uses your logo and lies about it, and that is squarely illegal, and you can win. A dupe copies your silhouette, your colorway, your entire visual signature, and brags about doing it, and in the United States, where we still offer almost no protection for apparel design itself, that is mostly lawful. We have tried and failed for two decades to pass a design-protection statute, and the absence is now a strategic fact, not an oversight. Generative tools will only widen the gap, because the cost of producing a convincing lookalike is collapsing toward zero.

Which leaves the conclusion that the entire five-hundred-year history points toward. The mark was always the technology of trust, and the lawsuit was only ever the defense of last resort. When the copy is illegal, you have a remedy. When the copy is merely permitted, you have only the thing a dupe cannot photocopy.

You have the brand. In the end, it was always the brand, which can never be duped.

The challenger has 30 days to appeal, and ballots print June 28, so keep an eye out for further developments!

There are, at the moment, two men named Dan Sullivan running for the same United States Senate seat in Alaska.

One is the incumbent, Dan S. Sullivan, a Republican who has held the seat since 2015 and is now seeking a third term. The other is Dan J. Sullivan, a former elementary schoolteacher and Forest Service employee from Petersburg, a town of about three thousand people on an island in the southeast panhandle. Both are running as Republicans. Both will appear on the August primary ballot, distinguished only by a middle initial. And the incumbent is, to put it mildly, not delighted.

He has called his namesake a “dirty political trick,” accused Democrats and the campaign of his likely rival in the general election of planting him to siphon votes under Alaska’s ranked choice system, and his party’s national committee has formally asked state officials to strike the newcomer from the ballot as a “sham.” There is even an allegation that the challenger borrowed the senator’s blue and gold campaign palette.

I have spent the better part of two decades helping people protect their names, their marks, and the identities they have spent years building. So when a story like this lands, I get the same question from three different directions within about an hour: Isn’t this a trademark problem? Can’t the senator just stop him?

The honest answer is one of my favorite kinds of answer, because it is the kind that is wrong in an interesting way. Mostly, no. And the “mostly” is where all the good law lives.

Your name is not a trademark (usually)

Start with the instinct, because the instinct is reasonable. “Dan Sullivan” is, in the context of Alaska politics, a recognizable identifier. People associate it with a specific person, a specific record, a specific blue and gold yard sign. That feels like a brand. Why shouldn’t it be protected like one?

Because trademark law was built to prevent a very specific harm: consumer confusion about the source of goods and services in the marketplace. A mark is not a reward for being famous. It is a tool for telling buyers who made the thing they are buying. Personal names get trademark protection only when they have acquired what we call secondary meaning, the point at which the public has come to understand the name as identifying a commercial source rather than just a guy. McDonald’s. Ralph Lauren. Even then, the protection runs against commercial uses.

A name on a ballot is not a good for sale. An election is not a marketplace in the Lanham Act’s sense, and a vote is not a purchase. Layer the First Amendment on top, where political speech and association sit at the absolute core of what the Constitution protects, and you can see why the law is deeply reluctant to let one candidate own a name and use it to evict another candidate who happens to share it. If we let incumbents trademark their way to unopposed ballots, we would have invented a powerful new tool for entrenching power, and courts know it.

So the senator cannot register DAN SULLIVAN and use it to remove Dan Sullivan. That is not how any of this works.

Which would be the end of a much shorter essay, except that it is not true that trademark law never touches politics. It does. Just not where most people expect.

Except when it is

The foundational case here is United We Stand America, Inc. v. United We Stand, America New York, Inc., decided by the Second Circuit in 1997. “United We Stand America” was the service mark of the political organization that grew out of Ross Perot’s 1992 presidential run. When a breakaway New York faction kept using the name, the national organization sued for trademark infringement, and won.

The defendants made exactly the arguments you would expect: our activities are political, not commercial “services”; our use is local, not interstate; and anyway, this is political speech protected by the First Amendment. The court rejected all three. Political organizing, it held, can absolutely be a “service” under the Lanham Act, and using another group’s name to identify yourself as part of their movement is the kind of source confusion the statute exists to stop. It was not writing on a blank slate, either. Courts had already applied trademark principles to advocacy groups soliciting donations, holding press conferences, and organizing on behalf of members.

So the line is not “politics is exempt.” The line is about what the name is doing. United We Stand America was a name that identified an organization and its services. The faction was free riding on that identity. That is a trademark harm, and the political setting did not launder it.

The slogan and the song

Two more episodes round out the picture, and both are instructive precisely because they cut in opposite directions.

In 2008, the McCain campaign, by way of an Ohio Republican Party web ad, used Jackson Browne’s “Running on Empty” without permission to mock Barack Obama’s energy policy. Browne sued. His copyright claim was the obvious one, but the cleverer claim was under the Lanham Act: by deploying a song indelibly associated with him, the ad falsely implied that Browne, a committed liberal, endorsed McCain. He added a claim under California’s right of publicity for the use of his voice. The court let the case proceed, and it ultimately settled, with the Republican Party issuing a public apology and a pledge to license musicians’ work going forward. False endorsement is a real lever in politics. When a campaign borrows an identity in a way that suggests a tie that does not exist, the law has something to say.

Now the opposite result, and it is hot off the press. In the 2026 California governor’s race, one candidate sued a rival for trademark infringement over a campaign slogan: “proven problem solver.” Last December, a federal judge threw it out. The phrase, he held, is generic political language, exactly the kind of cliché no candidate gets to monopolize, and he warned, rightly, against handing anyone a trademark over common political speech. He acknowledged the United We Stand and Browne lines and distinguished them. A name that identifies an organization is one thing. A platitude every third candidate prints on a banner is another.

Put those two next to each other and the principle snaps into focus. The law protects uses of an identity that point to a source. It does not protect the commons of ordinary political language, and it will not be conscripted to give incumbents a club.

The Jack Daniel’s hinge

If you want the doctrinal reason these cases line up the way they do, it runs through an unlikely place: a 2023 Supreme Court fight about a dog toy shaped like a whiskey bottle.

In Jack Daniel’s Properties v. VIP Products, the Court clarified when the usual tests flavored by the First Amendment give way to ordinary trademark analysis. The answer: when the defendant is using the challenged mark as a source identifier, as a brand, to say this is who I am, rather than purely as expression or commentary. That “source identifier” question is now the hinge. And it has already swung into politics. In 2024, the Sixth Circuit used exactly this reasoning in a fight between factions of the Libertarian Party over who got to use the party’s name and marks, holding that trademark liability can attach in the political arena when a name is being used to identify a source.

Which tells you how a court would actually frame the Alaska problem if it ever got there. The question would not be “is Dan Sullivan a famous name?” It would be “is the challenger using the name as a source identifier, to pass himself off as the incumbent’s enterprise, or is he simply a man named Dan Sullivan running for office?” And the law strongly presumes the latter, because the alternative is letting the state, or an opponent, decide who is allowed to use their own name in the most protected speech act there is.

The real tradition: election law, not trademark

Here is the part I find clarifying. The candidate who shares a name is not a trademark problem without precedent. It is an election law problem with a long and slightly absurd history.

In 1932, a man named Will Rogers, not the beloved humorist but a fellow who happened to share the name, was elected to Congress from Oklahoma, by most accounts on the strength of that name alone. The “ringer” or “decoy” candidate, deliberately recruited to confuse or split a vote, is a documented dirty trick that has surfaced in St. Louis, in Michigan (where one John Scott faced an election fraud charge for filing against an incumbent John Scott), and in plenty of municipal races where the strategy is cheap and occasionally effective. Alaska itself has been here before in spirit: the 2010 Murkowski campaign, run on votes written in by hand, turned partly on the mundane worry of how election workers would credit ballots among similarly named candidates.

And that is exactly how the Alaska fight will be resolved, not in a trademark court but through the machinery of election administration. The state’s lieutenant governor, who oversees elections, has already written to the challenger demanding he clarify his background and intentions; the relevant question is whether his ballot designation runs afoul of state rules against misleading or confusing presentations. Middle initials get added. Affiliations get scrutinized. Maybe he stays on the ballot, maybe he does not. But the tool is election law, where the state has a legitimate interest in orderly ballots that do not deceive. It is not trademark, where the senator would have to prove a commercial harm the Constitution is built to prevent him from manufacturing.

So what is protectable in Alaska?

If I were advising the senator’s camp, and to be clear I am not, I would tell them the name is the weakest part of their case. The name is a coincidence, and coincidences are not infringement.

The blue and gold, though? Now you are talking.

A consistent, distinctive visual identity, a color palette, a layout, a look that voters have come to associate with one campaign, is the closest thing in this whole story to a real trade dress and unfair competition theory. If the challenger truly copied the incumbent’s established design language in order to trade on the recognition it carries, that is a deliberate act of borrowed identity, not an accident of the baptismal font. It is still an uphill climb in the political arena, and the First Amendment still looms. But it is a real theory, where “he has my name” is not.

That is the whole lesson. A name is not an asset. Using it to point at a source is. Trademark law protects against confusion about who made this in the stream of commerce. It does not protect against the universe’s habit of producing two of you. The thing worth guarding was never the four syllables on the sign. It was the years of meaning poured into the look, the feel, and the trust that the sign points back to you.

There are two Dan Sullivans on that ballot. Only one of them has spent a decade building something a court could recognize. And, awkwardly for him, the part worth protecting is not his name.

If you build brands, run for things, or just enjoy watching the law trip over the messiness of real life, this is the kind of collision I write about often. Reply and tell me which shared name election I forgot. I know there are a hundred more.

That gap, between the people who create the language of a championship and the people who own it, is one of the most reliable patterns in American sports. Titles don’t just produce banners. They produce *words*. Chants, nicknames, slogans, logos. And trademark law, for all its reputation as a dry corner of commercial practice, turns out to be the machinery that decides who gets paid when a city falls in love with a phrase.

The law’s answer is almost never the romantic one. It doesn’t reward the person who said it first. It doesn’t reward the person it’s about. It rewards the person who files, who uses it in commerce, who keeps the paper current. The whole drama of sports IP is the distance between cultural authorship and legal ownership, and the NBA has been staging that drama for forty years.

## The patron saint

Start where every version of this story starts: with Pat Riley and the word “three-peat.”

Riley didn’t coin it. The credit belongs to his Lakers players. Byron Scott and Wes Matthews are both named in the lore, riffing on what a third straight crown would feel like after the Lakers repeated as champions in 1988. Riley did something none of them thought to do. In November 1988, he filed to register the term through his company, Riles & Co.

The Lakers promptly lost the 1989 Finals to the Pistons. Riley never three-peated, then or ever. It made no difference. When Michael Jordan’s Bulls won three in a row from 1991 to 1993, anyone who wanted to print the word on a hat or a shirt had to license it. From Riley. ESPN later estimated his take from the Bulls’ two three-peat runs at around $900,000. He collected again off the Yankees, off the Lakers’ own run in the early 2000s, and as recently as 2025 he cut a deal with the NFL for the Chiefs’ attempt. Riley once described the trademark as essentially free money lying on the sidewalk, and he wasn’t wrong.

Here’s the part New York should sit with. During that first Bulls three-peat, Pat Riley was the head coach of the New York Knicks, the team Jordan kept eliminating on his way to those very titles. Riley couldn’t beat the dynasty on the floor. He owned the word that described it. There is no cleaner illustration of the principle: on the court, you win by being best; in the registry, you win by being first.

One footnote that delights me as a practitioner. Riley’s original registration lapsed in 2012 because the renewal paperwork wasn’t filed, right as his own Heat were chasing a three-peat they would ultimately miss. He refiled and rebuilt the portfolio. Even the patron saint of foresight can drop a maintenance deadline. The system is unforgiving by design.

## The squatters and the star

Fast forward to February 2012 and a few miraculous weeks at the Garden. An undrafted point guard who had already been cut twice, Jeremy Lin, caught fire, and the city minted a word: Linsanity.

Within days, the filings started, and the first ones were not Lin’s. A man in California with no connection to the Knicks applied. Then another. By the time it was over, something like seven applications were in. The phrase itself had arguably been used in commerce earlier still, by a high school coach who’d bought linsanity.com back in 2010 and printed shirts.

Lin filed on February 13, paid his fee, and waited. And here the law did something closer to justice: the USPTO ultimately registered “Linsanity” to Lin. The office has tools to deny applications that falsely suggest a connection to a living person. You generally cannot register someone’s identity to cash in on their fame without consent. That gave Jeremy Lin an edge over the opportunists, even the ones who had filed first. Being first to the registry is powerful. It is not, quite, everything. When a mark *is* a specific person, that person holds an advantage no squatter can manufacture.

## The fan

Which brings us back to the present, and to two words that came from no athlete at all.

“Bing Bong” was born in October 2021, when a Knicks fan named Jordie Bloom hollered it into a Sidetalk street interview camera after a win on opening night, riffing on the chime of the closing subway doors. It went everywhere, the broadcast, the headlines, the team’s own social feeds, in under a week. By now you can see the shape of it. The apparel applications followed almost immediately, filed by entities with no obvious tie to the fan, the filmmakers, or the franchise.

This one is the hardest of all to “own,” because its meaning was authored collectively: a fan, a media account, a broadcaster, and a city, all at once. No single person made “Bing Bong” valuable, which is exactly why no single applicant has a clean claim to it. My favorite detail is that the oldest live federal registration for the literal words “BING BONG” has nothing to do with the Knicks at all. It covers *beer pong tables*, and it’s been on the books since the 2000s. The registry is not a map of cultural meaning. It’s a record of who showed up with a form.

## The logo

Trademark’s close cousin, copyright, produces an even sharper version of the lesson, and the NBA has the perfect case.

Kawhi Leonard is “The Klaw.” The logo that became his identity, a traced handprint hiding his initials and his number, started as a sketch he made in college. He shared it with Nike during his endorsement years; Nike’s designers built it out, and Nike registered the copyright. When Leonard left for New Balance, he sued, arguing the logo was his creation and that Nike had defrauded the Copyright Office. He lost. A federal judge held that Nike’s finished design was a new and distinct work, not merely a derivative of Kawhi’s sketch, and that the endorsement contract had vested the rights in Nike anyway.

Leonard won a title and Finals MVP with the Raptors in 2019 while that fight was pending, and a ring with the Spurs before it. The logo is *about* him; you cannot look at it and think of anyone else. None of that made it his. Association is not authorship, and authorship you’ve signed away is not ownership. Read your endorsement deals.

## The global game

Scale all of this up to the biggest brand the sport ever produced, and you get Michael Jordan in China.

Jordan is the championship brand: six titles, the dynasty whose three-peats funded Riley’s royalties. And for years he could not protect his own name in the world’s largest market. A company called Qiaodan Sports had registered the Chinese characters by which Jordan is known there, along with a logo of a jumping silhouette and the number 23, and built thousands of stores on it. Jordan sued in 2012 and lost dozens of times in the lower courts, because China rewards the first to register, not the first to be famous. Only in 2016 did the Supreme People’s Court hand him rights to his name in Chinese characters, while letting the company keep the romanized “Qiaodan,” because those registrations were old enough to be effectively untouchable.

The lesson for any brand with global ambitions is brutal and simple: file early, file defensively, file in jurisdictions where you haven’t sold a thing yet. Fame does not travel with a passport. Paperwork does.

## What the banner doesn’t tell you

Pull these together and a single rule emerges, stretching from Riley’s 1988 filing to whatever is being typed into the USPTO portal during this Knicks run: the championship creates the value, but the value flows to whoever did the unglamorous work of claiming it.

The player coins the phrase; the coach registers it. The city invents the chant; a stranger files for the merchandise. The athlete sketches the logo; the brand owns the copyright. The icon is beloved everywhere; the knockoff owns his name abroad. Every one of these stories is a small tragedy of authorship, and every one was avoidable with foresight, a filing, and a calendar reminder for the renewal.

I tell founders and brands the same thing every week, and the NBA has been proving it on a hardwood stage for four decades: cultural value and legal ownership are different things, and they do not automatically find each other. The time to protect an identity is *before* it becomes the thing everyone is repeating. By the time it’s a chant in the Garden, you are not the only one who wants it.

So enjoy the run. Scream the words. Just know that somewhere, someone is already filing them.

Bing bong. 🏀

I have learned to admire the baby and then ask the only question that matters at that stage, which is whether somebody else already owns it.

This is an awkward question, because the romance runs in one direction. You have fallen for the name. The market has not. To everyone outside the room the name is still a stranger, and a stranger is the one thing in branding you can still send away for almost nothing. Change a name before the logo, the packaging, the launch film, the press list, and you have lost an afternoon. Change it after, and you have lost a year and a number with too many zeroes in it. The whole discipline of trademark clearance lives inside that narrow window, the weeks when the name is still cheap to lose.

Here is how clearance has worked for most of my career. A company settles on a name it loves. A lawyer, if the company is careful enough to call one, orders a search. The firm itself almost never runs that search. It goes out to a vendor, one of a small number of companies that maintain the databases and run comprehensive screens across the federal and state registers, common-law uses, domain names, and the foreign registries that happen to matter for that business. The vendor sends back a report. The report is not a verdict. It is raw material, and it can run to hundreds of pages of marks that resemble yours in ways a database finds interesting and a human often does not.

Then the lawyer reads it. This is the part nobody outside the profession ever sees. Someone has to sit with those hundreds of pages and decide what is noise and what is a problem, which similar mark lives in a distant industry and which one sits close enough to ruin you, whether a dead registration is truly dead. Out of that comes a clearance opinion, a written judgment that says, in effect, you can probably use this, or you cannot, and here is why. The search itself can cost a thousand dollars or more. The opinion that makes the search mean something often runs two to three thousand, and sometimes well beyond it.

A good opinion does a second job, too, one founders rarely think about until it is too late. If you go ahead on the strength of a written opinion and someone later comes after you, that opinion is evidence that you acted in good faith. Good faith is frequently what separates ordinary damages from the enhanced damages courts reserve for infringers who knew better. The opinion is both the green light and, if you are ever challenged, the thing you hold up in your defense.

The strange part, once you have run a few hundred of these, is how much of clearance is machinery wearing a robe. The profession does not like to say this out loud. Law is one of the most information-dense trades there is, and one of the slowest to let a machine near the information, and the reason is not only temperament. It is arithmetic. For roughly a century the work has been priced by the hour, which means efficiency is the one improvement a firm is structurally punished for delivering. A tool that turns six hours into one is, on the old math, a tool that cuts your own invoice by five-sixths. Faced with that incentive, the profession did what anyone would do. It kept billing the six hours.

Clearance is the cleanest example of this I know. Strip the task down and most of it is pattern matching at scale, the comparing of one proposed name against millions of registered and unregistered uses, exactly the sort of labor software has been good at for years. The judgment sits on top, a thin and crucial layer where a trained lawyer decides what the matches mean. For decades the thin layer and the thick one were bundled together and billed together, by the hour, so the part a machine could have done cheaply was sold at the price of the part only a person can do. The client paid lawyer rates for database work because the database work arrived stapled to a lawyer.

The first technology that tried to unbundle any of this reached consumers, not firms. The document-automation companies, the fill-in-the-blanks generation of legal products, took the cheap, repeatable parts of legal work and sold them cheaply and directly. For a simple will or a single-member LLC, that was a real gift. For anything where being wrong is catastrophic rather than inconvenient, it was a trap dressed as a bargain, because it handed people the mechanical output and removed the judgment that was the only reason to do the exercise at all. Trademark clearance sits firmly in the catastrophic column. A will done slightly wrong can often be fixed after the fact. A name adopted slightly wrong takes the whole company down with it.

The current wave is more capable than anything before it and dangerous in a new and specific way. It is fluent. The earlier tools at least looked like software, with their forms and their checkboxes and their disclaimers, and a user knew to hold them at arm’s length. A large language model writes back in clean, confident prose, and fluency reads as authority. When one of these systems invents a registration that does not exist, or fails to surface one that does, it does so in the same untroubled voice it uses when it happens to be right. A confidently wrong clearance is worse than no clearance at all, because no clearance at least leaves you nervous. A wrong one sells you peace of mind you have not earned, and then you build on it.

Now do the math a real founder does. You do not have one name you love. You have ten. You want to know which of the ten will survive contact with the world. At clearance rates, screening all ten is a five-figure exercise before you have sold a single unit, so most founders do the apparently rational thing and skip it. They pick the name they like best, they build, and they hope.

Hope is not a search. When we trace an infringement suit back to its origin, the origin is almost always the same. The name was never cleared. Sometimes nobody looked. Sometimes somebody looked, saw something they did not like, and talked themselves out of it. Either way the company is now spending on lawyers and a rebrand a multiple of what a clearance opinion would have cost, and it is doing so while the business sits inside the interruption of a name it can no longer use. This is not a rare event we have read about somewhere. It is most of what we do.

Consider Rhode. In 2014 two former roommates, Purna Khatau and Phoebe Vickers, built a fashion label under that name and secured the federal trademark. In 2018 Hailey Bieber’s team approached them about buying it, which tells you the name was a stranger to no one involved. The roommates declined. In June of 2022 Bieber launched a skincare brand called Rhode anyway, and within days she was in the Southern District of New York defending it. She won the early round, when the court refused to shut the brand down while the case proceeded. The fight still ran two years before the parties settled on terms they kept to themselves. Bieber has the resources to absorb a thing like that. Almost none of the founders I meet do. And notice that this was not a failure to look. Somebody looked. The value a clearance opinion adds is not only discovery. It is the cold reading of what you found, delivered before you have spent a dollar on packaging and a documentary about the name.

Or consider the less public version, the one that ends in a new name. A London dating startup called 3nder, all of ten employees, built an audience around a word that rhymed a little too neatly with Tinder. Tinder’s owner sent a cease and desist, then sued. The founder described it as a giant aiming a weapon at a company that could fit in a single room. In 2016 they gave up the fight over the name and became Feeld, which is what they are still called. The product survived. The name did not. Every download, every mention, every scrap of recognition built under the old word had to be rebuilt under the new one, and that rebuilding is the part of a rebrand that never appears on an invoice and costs the most.

When we started building Thanks Mom, our trademark search and clearance product, we did the unglamorous thing and talked to general counsel, as many as would sit with us, and we are still doing it. We expected to hear that they wanted the technology, and they did. What surprised us was how uniformly they named the same hesitation. The tools exist, they would say, and some are impressive, but there is no one on the other end. No malpractice policy. No name on a door. No human being who answers for it when the software is wrong. You cannot sue a chatbot, and you certainly cannot ask one to stake its license on the answer. They were not willing to bet a brand on a hallucination, and they were right.

That hesitation is the whole game, and it is worth saying plainly what sits underneath it. A law firm is a thing that can be held responsible. It can be sued, it carries insurance, it answers to a licensing body, it holds a license it can lose. Those are not bureaucratic details. They are the mechanism by which a client moves risk off its own books and onto someone who has promised, under real consequence, to be right. An automated tool, however good, cannot accept that transfer. It has nothing to lose. Accountability is not a feature you bolt onto legal software. For the decisions that actually matter, it is the product.

So we built Thanks Mom around a refusal to make anyone choose between the two. The product uses the best search technology available, runs the screen, and drafts the clearance opinion. Then it stops. Nothing reaches the client until one of our lawyers, an actual person at our firm, has read the work and stood behind it, the same way a law firm stands behind any opinion that leaves the building. We keep a human in the loop, in this product and in everything that carries the firm’s name. What the client receives has the speed and the cost curve of the machine and the accountability of a law firm. You get the best of the technology and the best of the people, and you never have to guess which one was right.

This is the part of the legal-technology conversation I find most misunderstood. It is almost always framed as a contest, the lawyer or the machine, as though one of them has to win. The framing is wrong. The work has layers, and the layers want different things. The bottom layer, the volume, the searching and sorting and the first rough draft, is labor the machine should have taken over years ago and can now do beautifully. The top layer, the judgment and the answerability, has to stay with a person, because a person is the only party who can be held to it. The firms that come out ahead will not be the ones that automate everything, or the ones that automate nothing out of pride. They will be the ones that put the machine under the lawyer rather than in front of the client, and let each do the thing it is genuinely good at.

We built Thanks Mom for brands, but the people who understood it fastest were other lawyers, and that taught us something we should have seen sooner. Every trademark lawyer in the country is running the same expensive, manual process we were, out of sight of the client. Order the search. Read the report. Write the opinion. Bill two or three thousand dollars. The work is necessary and the price has been fair, but most of that price is hours, and most of those hours buy the parts a machine now handles well. So we are building a second version of the product, this one made for attorneys.

The arithmetic for a lawyer is hard to argue with. Run a clearance through it and you can produce the same opinion, to the same standard, with your own judgment still sitting on top, at roughly a quarter of what the work costs you to produce today. From there the lawyer has two choices, and both of them are good. Keep your rate where it is and keep the difference, which turns clearance from a grudging, thin line item into one of the better margins on the menu. Or pass the savings through, clear a client’s ten candidate names for something close to what one name used to cost, and become the lawyer who made clearance affordable rather than the one who made it hurt. One path feeds the firm. The other delights the client. Most of the lawyers I know will do some of each, depending on the matter and the relationship, and they will be right to.

What is not on the table, for the lawyer who picks this up, is being replaced by it. The judgment they bring is the same judgment as before. What changes is the leverage underneath it, the ratio of result to hours, and that ratio is the whole future of a firm like mine. I have spent years arguing, to anyone who would sit still, that the next era does not belong to the firms with the most bodies. It belongs to the firms with the most leverage per body, the ones that can do the work of a much larger shop without becoming one, and stay small enough to still be good. A tool that lets a single trademark lawyer clear names with the throughput of a department, and personally answer for every one of them, is exactly what that future is built from.

There is another consequence I care about more than the margins. When clearance costs five figures, only the well-funded clear their names (or get adventurous trying). Everyone else, the first-time founder, the artist turning a project into a label, the small brand with more taste than capital, does the math and gambles, and a disproportionate share of the people who get sued over a name are simply the people who could not afford to check. Cut the cost of doing it right by three-quarters and you have not only widened a few margins. You have let a whole tier of people who were priced out of caution begin to care, properly, about something they should have been able to care about all along. I find that a better reason to build the thing than anything on a spreadsheet.

The romance of naming is real, and I would not talk anyone out of it. The name is often the first true thing a company makes, the place where the whole idea finally takes a shape you can say out loud. I only ask founders to do the romance in the right order. Clear the name while it is still a stranger and still cheap to lose. The difference now, after years of the work being priced out of reach for most of the people who needed it, is that clearing it no longer has to be the expensive part. Do that first. Then go ahead and fall in love. By the time you do, it will be yours to keep.

That moment, multiplied by ten million across TikTok, Amazon, Temu, Shein, and the cookie aisle at Aldi, is one of the hottest issues moving through consumer packaged goods right now. It does not involve a celebrity defendant or a viral cease-and-desist or a Supreme Court oral argument with audio that lawyers email to each other. It involves something a little quieter, but probably even more consequential. Brand owners have woken up to the fact that the most valuable real estate they own is no longer the name on the front of the package. It is the package itself. And they are racing, all at once, to protect it.

The legal term for this is trade dress. It is a phrase nobody outside our profession enjoys saying out loud, and I will keep its use to a minimum, but it is worth defining once, plainly. Trade dress protects the look of a product or its packaging, the things that make you reach for one item on a shelf instead of the one next to it. The fluted contour of a Coca-Cola bottle. The orange of a Tide jug. The little foil-wrapped triangle of a Hershey’s Kiss with its paper plume sticking out the top. None of these are words. None of them are logos. They are shapes and colors and silhouettes, and for most of the twentieth century they sat quietly in the background of American commerce while lawyers fought about word marks and slogans, which were easier to draw and easier to litigate.

That is changing very quickly, and the reason it is changing is not legal but cultural.

## A short history of looking like something

The doctrine itself has a long, slightly creaky history. For most of the past century, trade dress was treated as a footnote to trademark law. It existed in case law and in section 43(a) of the Lanham Act, but it lived mostly on the margins, invoked when a competitor copied the décor of a restaurant or the labeling of a bottle. The first time the Supreme Court took it seriously as a free-standing doctrine was 1992, in a case called *Two Pesos v. Taco Cabana*, which I find I have to keep explaining to non-lawyers because the facts sound made up. Two chains of Mexican fast-casual restaurants in Texas. One sued the other for copying its “festive eating atmosphere,” a phrase the court used without irony, complete with bright colors, murals, awnings, and what the trial record describes as a stepped exterior in vivid paint. The Supreme Court held that a restaurant’s overall look and feel could be protected as trade dress, and that if that look was inherently distinctive, the plaintiff did not have to prove what we lawyers call secondary meaning. It was a quiet ruling that opened a wide door.

Eight years later the Court walked some of it back. In *Wal-Mart v. Samara Brothers*, decided in 2000, Justice Scalia, writing for a unanimous Court, held that the design of a product itself, as opposed to its packaging, can never be inherently distinctive. To protect the shape of your dress or your toy or your blender as trade dress, you have to show that consumers have come to associate that shape, specifically, with you. You have to show secondary meaning. The case involved a line of children’s clothes with appliquéd hearts and flowers that Wal-Mart had asked a contractor to reproduce for its own racks, and although the underlying dispute was small, the doctrinal split it created has shaped trade dress strategy ever since. Packaging is one thing. The product itself is another, and harder.

For most of the next two decades this distinction sat where most doctrinal distinctions sit, which is in casebooks and in the back of trademark attorneys’ minds. The cases that did make news tended to be the exotic ones. The red sole of a Louboutin shoe. The shape of a Hermès Birkin bag. The interior of an Apple store, which the company successfully registered in 2013. These were luxury cases, fashion-house cases, the kinds of disputes that produced glossy law review articles and very few practical consequences for brand owners outside of fashion and culture.

And then dupe culture exploded.

## The shelf, and what is on it

I do not know exactly when “dupe” became a verb, but I can tell you when it became a market. Sometime around 2022, the algorithm started rewarding side-by-side comparisons. A teenage girl in a bathroom mirror, two perfume bottles in her hand, swearing one of them smelled exactly like the other. A Costco run with the camera lingering on a Kirkland Signature label next to a hashtag for the brand it was almost. Run, don’t walk, the captions said. It worked because consumers wanted to believe they had cracked a code, that they had found the same thing for less, and because the platforms wanted them to keep believing it. By 2024 the hashtag economy around dupes was generating billions of views a month. By 2025 it was generating lawsuits.

The Lululemon case, filed in the Central District of California in June 2025, was the one that signaled the shift. Lululemon sued Costco for trade dress infringement, alleging that the Kirkland Signature versions of its bestselling jackets, sweatshirts, and pants were not just inspired by Lululemon’s designs but engineered to be mistaken for them, complete with influencer marketing on TikTok that explicitly used the word “dupe” as a selling point. The complaint named the #LuluDupes hashtag as evidence of confusion. What was interesting about the suit was not the legal theory, which was a fairly standard trade dress and unfair competition claim, but the strategic posture. A premium brand was treating the dupe economy as an existential threat and going to federal court to say so.

Lululemon and Costco Kirkland Signature leggings shown side by side in a viral dupe comparison.

A month earlier, in May 2025, Mondelez had filed a similar but broader suit against Aldi in the Northern District of Illinois. Mondelez, through its subsidiary Intercontinental Great Brands, owns most of the iconic American snack aisle. Oreo. Chips Ahoy. Ritz. Nutter Butter. Wheat Thins. Premium. Nilla Wafers. The complaint alleged that Aldi’s private-label versions of all of these had been deliberately designed to mimic the Mondelez packaging, down to the color blocking, the font weight, and the angle at which the cookie or cracker was photographed on the front of the box. The filing included page after page of side-by-side images. The Aldi blue next to the Oreo blue. The Aldi red next to the Chips Ahoy red. The Mondelez complaint also included a striking detail, which is that this was not a first offense. According to the filing, Mondelez had been sending cease-and-desist letters to Aldi for years, about copycats of Oreos, Teddy Grahams, Belvita biscuits, Triscuits, and Tate’s Bake Shop cookies, and Aldi had sometimes complied and sometimes not. The 2025 lawsuit was the moment Mondelez decided the letter-writing era was over.

By the end of 2025 the dam had broken. Sol de Janeiro sued the Australian brand MCoBeauty for copying the look of its Cheirosa body mists, the elongated transparent bottles with the numbered ovals and the pastel caps that any teenager who has been in a Sephora in the last three years can picture without effort. The suit alleged not only trade dress infringement but false advertising, citing influencer videos in which MCoBeauty’s products were promoted as smelling “exactly like” Sol de Janeiro’s. MCoBeauty moved to dismiss in early 2026, arguing that the packaging elements were functional, that the comparisons were puffery, and that dupe shopping is lawful competition rather than confusion. The motion is pending as of this writing. Whoop sued a Chinese seller on Amazon called Shenzhen, alleging that wearable trackers sold under names like SGJIK and EGQINR were trading on its trade dress. Deckers sued Costco over a slipper that looked too much like an Ugg. Smucker’s sued an Aldi look-alike of Uncrustables.

There is something almost archaeological about reading these complaints in sequence. You can watch the lawyers, ours included, learning to speak a new language. Phrases that used to belong to luxury fashion litigation are now appearing in cracker and body-mist filings. *Color-coded ovals*. *Sans serif numerals*. *Pearlescent caps*. *Overall commercial impression*. The taxonomy of a brand’s appearance, which used to live in a marketing deck, is now being translated into something a federal judge can grant an injunction over.

## Why now

A reasonable question, and one a client asked me last week, is why all of this is happening at once. These trademarks, for the most part, have existed for decades. The contour Coca-Cola bottle has been a registered mark since the 1960s. The Hershey’s Kiss shape has been protected for almost as long. The Oreo cookie itself, with its raised flower pattern and its precise dimensions, has been recognizable since 1912. The packaging for Ritz crackers, with the red and the gold and the cracker fanned out diagonally, has been on grocery shelves since the Roosevelt administration. So why the flurry of new applications in 2026, the cease-and-desists going out by the dozen, the litigation calendar suddenly crowded with disputes about the angle of a font and the precise saturation of a blue?

The honest answer is that the threat changed shape. For most of the twentieth century, a knock-off was a regional problem. A bakery in Ohio made a cookie that looked like a Ritz. A small toy company in California made a doll that looked like a Barbie. The Mondelezes and Mattels of the world sent letters, filed the occasional suit, and went on with their lives. The economics of copying simply did not scale. You needed real distribution to compete with a national brand, and real distribution meant real exposure, and real exposure meant you were findable, suable, and stoppable. Distribution was seemingly almost as valuable as the IP.

That world is gone. The combination of e-commerce, social platforms, drop-shipping logistics, and large-format discount retailers has produced an environment in which a copy can reach millions of consumers in weeks, with no factory, no contract, and no visible owner. A seller on Amazon under a randomized string of capital letters can move tens of thousands of units of a knock off fitness tracker before the original brand’s general counsel has finished her coffee. A TikTok creator can route an entire generation of teenage shoppers to a body mist that smells “exactly like” the one they actually wanted, and the platform’s response, when the brand owner complains, is often a form email. Aldi, Costco, Trader Joe’s, and the regional discount chains that have grown into national powers in the last decade have made the private-label aisle into a kind of permanent shadow market, and the shadow is now darker than the thing casting it. Aldi opened more than two hundred new stores in the United States in 2025 alone. Its slogan, displayed on the bag every shopper carries home, is *Like brands, only cheaper*. The lawsuit Mondelez filed reads, in places, like a forensic decoding of exactly what that slogan means.

Brand owners are responding the only way they know how, which is by going to the United States Patent and Trademark Office and trying to lock down what they should have locked down years ago. The filings have accelerated noticeably in 2026. In the last few months alone, Intercontinental Great Brands has filed or refreshed trademark applications covering the packaging of Animal Crackers, Ritz, and the visual appearance of Swedish Fish, the candy itself, the little red fish with the matte finish that has been in American checkout aisles since 1957. These are not new products. They are not even new trademarks in any deep sense. They are old, recognizable, beloved bits of commerce that the company is suddenly, urgently, registering with new specificity, new claim language, new color drawings, new descriptions of which elements are functional and which are not. If you have ever wondered what corporate panic looks like when it is filtered through a USPTO filing portal, this is what it looks like. It looks like a fifty-year-old candy being re-papered as if it were brand new.

## What the doctrine actually does

It is worth being precise about what these filings can and cannot accomplish, because the gap between what brand owners want and what trade dress law gives them is one of the more interesting questions of the next several years.

If you can register the packaging of a product as trade dress, you get something close to what you get with a word mark. You get the presumption of validity. You get the right to sue in federal court without proving secondary meaning from scratch every time. You get a public record that puts competitors on notice. You get, in practical terms, leverage. A cease-and-desist with a registration number attached is a different kind of letter from one without. Aldi’s lawyers know this. Costco’s lawyers know this. The seller on Amazon hiding behind a string of capital letters knows this too, or rather, the platform’s takedown team knows it for them.

If you are trying to protect the product itself, the shape of the cookie or the silhouette of the legging, the law is harder. The Samara Brothers case still controls. You have to prove that consumers see your design and think of you, not just of the category. That proof is expensive. It requires surveys, advertising spend records, sales figures, and what we sometimes call “look for” advertising, which is exactly what it sounds like: ads that explicitly tell the consumer, “look for the diamond shape”, “look for the orange sole”, “look for the fluted bottle”. Brands that have been doing this all along have a real advantage. Brands that have been letting their product design speak for itself, on the theory that quality sells, are discovering that quality is exactly what their competitors are now able to imitate. The cookie is the cookie. The leggings are the leggings. What separates them is the story you have told about them, and trade dress law cares deeply about whether you have actually told that story in a way the consumer received.

There is also the functionality doctrine, which trips up almost everyone. You cannot use trade dress to protect a feature that exists because it works. The shape of a hammer head, the contour of an ergonomic mouse, the precise geometry of a bottle that pours without dripping. These belong to the patent system, which is temporary by design. Trade dress is forever, and the courts are vigilant about not letting brand owners use it to lock down what is really a functional advantage. Many of the dupe defenses you are seeing right now, including MCoBeauty’s, lean hard on this argument. The bottle is transparent because perfume bottles are transparent. The cap is rounded because caps are rounded. The font is sans serif because everyone uses sans serif. There is something to this, and there will be a body of case law over the next two years that tries to draw lines a federal judge can hold a ruler to.

## What changes in the boardroom

The interesting consequence of all this, for those of us who advise brand owners, is that the way companies think about IP has started to shift. For a long time, trademark strategy was the thing you delegated to outside counsel and reviewed once a year. You filed your word marks. You renewed your logos. You sent the occasional letter. The serious work, the work that absorbed attention at the executive level, was patent strategy and copyright licensing and the deal book. Trademarks were the table stakes.

That is no longer true at the companies I work with. The conversation has moved into the C-suite. General counsel are being asked to map every visible element of every product the company sells, to flag which are registered, which are unregistered, which could be registered with effort, and which are too functional or too generic to protect at all. They are being asked to coordinate with marketing, which has historically been the first to know when a competitor has copied a campaign and the last to be consulted about the legal consequences. They are being asked to build what one client of mine has started calling a “look book,” which is a polished internal document, updated quarterly, of every distinctive visual element the company owns or wishes it owned. It reads more like a brand bible than a legal filing, and that is the point. The point is that the look and the law have to live in the same document now, because the threats are arriving through both at once.

The other shift is in enforcement. For the last decade, the cease and desist letter has been the default tool. Send the letter, hope the recipient backs down, and if not, escalate. The dupe era has made the letter less useful, partly because the recipients are often anonymous and overseas, and partly because the platforms themselves have become the relevant audience. A brand owner with strong, registered trade dress can take down a TikTok video, an Amazon listing, a Temu storefront, a Shein product page, often within hours, without ever filing a complaint. The registration is the leverage. The takedown is the remedy. And the lawsuit, when it comes, is often as much a public statement as a private dispute. Lululemon’s complaint against Costco was, among other things, a press release. So was Mondelez’s complaint against Aldi. So was Sol de Janeiro’s against MCoBeauty. The legal filing has become a brand-protection asset in its own right.

## A small confession

I should say, before closing, that I am not entirely a neutral observer of any of this. Our firm represents fashion houses, iconic brands, and consumer products companies whose entire business depends on the proposition that the way a thing looks is the thing. We have spent years sitting in conference rooms staring at packaging and asking whether the green is the green and whether the silhouette is the silhouette, and we have been, on more than one occasion, the people sending the letter and the people receiving it. I know what it feels like to watch a client’s distinctive product appear, six months later, on Amazon under a name that looks like a license plate. I know what it feels like to be told by a platform that the takedown is denied. I know what it feels like to file the application five years too late.

What I am describing is, in part, the consequence of a long inattention. American brand owners spent most of the last quarter century treating trade dress as a luxury problem, something Hermès worried about. They are now discovering that the discount retailer at the end of the strip mall is a more sophisticated and patient copyist than any counterfeiter ever was, and that the platforms they relied on for distribution will not protect them by default. The scramble of 2026 is the scramble of an industry that has belatedly understood the asymmetry. The copyists move fast. The legal system moves slowly. The only way to be ready when the asymmetry breaks is to have the paperwork already done.

Which is why my desk, and the desks of every trademark lawyer I know, is suddenly stacked with applications for the appearance of products that have been on shelves my entire life. The Swedish Fish. The Ritz cracker. The little red fish, the little gold box. We are not protecting them because they are new. We are protecting them because everything around them has changed.

The shelf is the same. What is on it, and who is making it, no longer is.

When I started doing these deals in 2009, nobody called them collabs. They were trademark licenses. That is what the contracts said at the top, that is what the lawyers on the other side called them, and that is how the brands understood them. A trademark license, with the artist or the smaller brand as the licensor, the bigger brand as the licensee, and a set of clauses copied from a form some law firm had been using since the 1990s.

The deals were almost always shoehorned.

Shoehorned because the form that was used had been built for a world where a celebrity endorsed a fragrance or an athlete put his name on a sneaker, and the celebrity or the athlete walked away with a check and no further claim on what they had touched. Shoehorned because the brand assumed, by default, that any IP generated in the course of the partnership belonged to the brand. Shoehorned because the artist’s lawyer, if the artist had a lawyer, often did not know to push back, and the artist, if they did not have a lawyer, signed the thing and learned six months later that they had given away every drawing, every sketch, every secondary use right, every approval over the next product in the line.

I spent a lot of years explaining to brand counsel that the artist was going to retain their IP. It often felt like a tutorial! Some of those tutorials took three meetings. Some took six. The brands had never done a deal where the creative half of the room walked away owning anything, and the muscle memory of the boilerplate they were using fought it the whole way.

The brand-on-brand thing, the thing that gets called a collab now, did not really exist yet. There were antecedents, of course. Schiaparelli and Dalí in the 1930s. Stephen Sprouse painting graffiti on LV bags in 2001. Murakami on the LV monogram in 2003, which most people in the industry now point to as the line that set the mold for collaborations between artists and luxury brands. But the cycle of brand X doing a capsule with brand Y, every season, across every category, with both names sharing space on the hangtag and a press release queued up for Hypebeast, that was not yet a category. The first wave was artist-meets-brand. The second wave, the brand-on-brand wave, the streetwear-meets-luxury wave, did not really hit until the early 2010s.

I have been there for most of it. The firm has been there for most of it. We have written hundreds of these deals across that arc, and watched the form evolve from a trademark license with an awkward graphics rider (lol iykyk) into a fully developed body of work with its own grammar, its own clauses, its own land mines, and its own jurisprudence.

What follows is the anatomy of that deal. The clauses that matter, in the order they tend to matter, with the collabs that taught the industry what each one is actually worth. I want to start with the one that taught me the most.

The Collab That Rewrote the Rulebook

Daniel Arsham x Tiffany & Co. is, in my view, the most instructive collab of the last decade.

I am biased here. I represent Daniel. I have watched this partnership grow from a single object in 2021 to maybe the longest-running, deepest creative collaborations in luxury today. But the reason it sits at the center of this essay has nothing to do with how it started. It has to do with what Tiffany was willing to do that no luxury house had done before, and what every clause of every collab agreement is ultimately trying to make possible.

The first piece was the Bronze Eroded Tiffany Blue Box, Circa 3021. A stunning life-size version was installed at the jeweler’s flagship store on Fifth Avenue in Manhattan, and visitors could step inside the sculpture and view an exhibition of the Tiffany Knot x Arsham Studio jewelry collection. A walk in artwork, in the most valuable retail real estate in American jewelry, treating the most recognizable piece of luxury packaging in the world as a future archaeological artifact. That alone would have been enough.

But the real story is the logo.

Tiffany & Co. has been Tiffany & Co. since 1837. The brand mark is one of the most protected, most policed, most sacred trademarks in luxury. And in the course of this partnership, Tiffany allowed Daniel’s interlocking double-A signature to share space with their own marks, on packaging, on co-branded boxes, on the lockup itself. First time that has happened in the history of the house. That is the kind of thing that, in many other luxury companies, gets killed in the first approvals meeting. The brand team says no. The trademark counsel says no. The CMO says we have never done that and we are not starting now.

Tiffany’s out of the box, adventurous thinkers said yes.

Then they said yes again, when the partnership extended to the Tiffany Knot, then the Tiffany Lock, then a bust based on the Venus of Arles, then a permanent Arsham sculpture inside the remodeled Fifth Avenue flagship. A bespoke Arsham statue graced the entrance of Tiffany’s historic Saatchi Gallery exhibit, and another is given permanent pride of place within Tiffany’s remodeled Fifth Avenue flagship.

Then they said yes to Pokémon.

The 2023 Tiffany & Arsham Studio & Pokémon capsule is the best example I have ever seen of a luxury house letting a creative partnership pull it into territory the house, on its own, would never have entered. The collection featured nine jewelry designs across six iconic Pokémon — Pikachu, Charmander, Squirtle, Jigglypuff, Cubone or Mew — with a Tiffany Blue Poké Ball housing the gold Pikachu pendants. A Tiffany Blue Pokéball. Read that sentence again.

It’s pretty remarkable all this happened in perfect sync. All built on trust.

That trust is the asset every collab is trying to build and almost none of them get to. It is not in the term sheet. It is the thing the term sheet exists to protect.

Now, the clauses that protect it.

The Marks

Every collab is, at its heart, a trademark license. A trademark license is permission, granted by the owner of a mark, for someone else to use that mark on specified goods, in specified territories, under specified quality controls, for a specified period of time. One party owns a mark. The other party wants to put that mark on a product (could be copyright, in the case of an artist). Everything else is plumbing.

The license has to specify exactly which marks (or copyrights), on exactly which goods, in exactly which territories, for exactly how long. The fight is almost always over the how. Can the licensee use the licensor’s primary logo, or only a co-branded lockup? Can they alter colorways? Can they use the mark on hangtags, in advertising, in social, in retail signage, on the box?

The 2017 Louis Vuitton x Supreme deal is the textbook here, and not because it was elegant. It was a one-season co-branded apparel and accessories run, sold through pop-ups, that put Supreme’s red box logo on top of Louis Vuitton’s monogram. In this deal, the luxury house, in this configuration, was the licensee of the streetwear brand’s mark, not the other way around. Every clause downstream of that decision had to follow the trademark logic. Whose quality control standards apply when the streetwear brand is the licensor? Whose approvals govern the lookbook? You write the deal differently when the small brand is calling the shots.

The first question in any collab term sheet is the boring one nobody wants to ask. Whose mark, on whose goods, controlled by whom. If you can’t answer that in one sentence, you don’t have a deal, or, as the contract lawyers always say, a “meeting of the minds.”

Approvals and Quality Control

Trademark law requires the mark owner to control the quality of the goods that bear the mark. This is not an aesthetic thing. It is the price of admission. A licensor who fails to police quality risks abandonment of the mark. Which means the approvals clause, the thing the creatives think of as bureaucratic friction, is actually the structural beam holding up the entire license.

In a healthy collab, the approvals matrix is granular. Materials, colorways, prototypes, packaging, hangtag copy, marketing creative, press release language, social posts, retail merchandising, product seeding, etc. Every category gets a named approver, a turnaround window, and a fallback for when the approver ghosts. The deal that gets signed without a fallback is the deal that ships three weeks late because someone’s CMO is on holiday in St. Tropez (it has happened!).

The Tiffany x Nike Air Force 1 collab is a fascinating case study in what happens when the approvals process produces something the market reads as too restrained. The 2023 release was a black suede AF1 with a Tiffany Blue Swoosh and silver heel accents, priced at $400. The internet wanted a Tiffany Blue shoe with diamonds on it. They got something much quieter. You can read that as a marketing miss, or you can read it as exactly what a 187-year-old jeweler’s brand approval process is designed to produce. Tiffany’s quality control instincts protected the mark. Whether it produced the product the culture wanted is a different conversation.

The clause does its job. The job is not always what Dimes Square wants!

Exclusivity and Category

Exclusivity is where collab deals get expensive, fast. Exclusivity is the clause that says who else the licensor can do a deal like this with, in what category, against which competitors, and for how long.

The licensor wants a narrow grant. Footwear, only Air Force 1, only this colorway, only this co-branded lockup. The licensee wants a wide grant, with options. Sneakers and apparel and accessories, with a right of first refusal on the next category, in the territory of the world, exclusive against direct competitors for two years post-launch.

The negotiation lives in the modifiers. Exclusive within sneakers is not the same as exclusive within footwear, and exclusive within footwear is not the same as exclusive against Brand X by name. I have seen deals where a brand thought it had locked up its partner, only to discover that the partner was free to do an identical capsule with a competitor six months later because the exclusivity language was scoped to a SKU type that did not technically include the competitor’s product.

You write exclusivity by category, by competitor, and by time, and you say what happens when any of the three drift. People working on this part of the deal should feel free to be as detailed and thorough as possible - they’ll be happy later they did.

Term and Sell-Through

The term is the runway. The sell-through is the landing. Term is how long the license lasts, and sell through is the winddown period after the term ends, during which the licensee can keep selling existing inventory but cannot make any more.

Most collabs run six to twenty-four months from launch. The complicated part is what happens at the end. The brand has manufactured product. Some of it has shipped. Some of it is in DCs. Some of it is in retail. Some of it is in returns. The license cannot just snap off, because the goods continue to exist in the world, and goods bearing both parties’ marks have to either be sold through, destroyed, or rebranded. This is a practical, real-world thing that has to be dealt with in the contracts.

The standard structure is a sell-through period of sixty to one hundred eighty days post-termination, during which the licensee can continue to sell existing inventory but cannot manufacture new units. Beyond that window, unsold inventory has to be either reacquired by the licensor at cost, destroyed with proof of destruction, or, in the more creative deals, marked down and sold through outlet channels with reduced or eliminated royalty obligations.

The Adidas Yeezy unwind is the case study every collab lawyer should read twice. Adidas terminated the partnership in October 2022. Adidas was left with $1.3 billion worth of unsold Yeezy sneakers, and chose to sell the stock and donate part of the proceeds to charity. The reason that decision was even available to them is that the deal had been written with an asymmetric IP structure. Adidas was in charge of design and manufacturing, and gained ownership of the designs as part of the deal, with relevant trademark, copyright, and patent registrations issued to and maintained by Adidas. Ye owned the Yeezy name. Adidas owned the silhouette. Which meant after termination, Adidas could sell the existing shoes without the Yeezy mark, because they owned the design itself. The contract may include termination clauses that allow one party to take the IP from the other, or a morals clause that allows early termination of the deal, and the Adidas-Ye deal had both.

The lesson is brutal. Who owns the design is a different question from who owns the brand, and a collab agreement that does not separate those questions cleanly leaves a billion dollars of inventory hanging in limbo. In other words, the name on the tag and the shape of the shoe are two different pieces of property, and if your contract does not say who owns which one, you will find out the hard way the day the partnership ends.

Economics

The money is rarely the hardest part to negotiate. It is almost always the hardest part to model.

There are four basic economic structures, and most collabs use a combination of them.

A flat fee, paid up front, in exchange for the license. Clean, simple, dumb. The licensor gets paid no matter what the deal does at retail. The licensee assumes all of the upside and all of the downside.

A royalty on net sales, usually expressed as a percentage of wholesale or net revenue. The Yeezy deal was famously written at 15% of royalties off of Yeezy sales, paid to Ye as the licensor of the brand to Adidas. That number is the high water mark of celebrity-brand collab economics, and it is the reason every artist who walks into a deal room asks for fifteen and gets told they are not Kanye.

A profit share, calculated post-COGS and post-marketing, which sounds great until you read the definition of deductible expenses or net profits and realize that the licensor is funding half of the licensee’s overhead.

And a guaranteed minimum, layered on top of any of the above, which is the licensor’s way of saying I do not trust your sales forecast, so pay me the floor regardless.

The structure you pick depends on who has leverage. A streetwear brand collabing up to a luxury house probably takes a flat fee plus a royalty kicker. A celebrity brand collabing down to a sportswear company takes the royalty and the marketing fee. The Yeezy deal famously included a marketing fee to promote the shoes — $51 million in 2020, or about 3% of net sales, on top of the royalty stream. That is the kind of structure you only get when the other side believes you can move the entire category.

Marketing, Messaging, and Approvals

The marketing clause is where some collabs fall apart.

A marketing clause specifies who runs the launch, who funds the spend, who approves the creative, what each party can say in earned media, what each party can say in paid media, and what happens when one party’s spokesperson says something off-message in a podcast interview.

The agreed language section is its own discipline. Press releases get pre-cleared. Talking points get pre-cleared. Influencer briefs get pre-cleared. The brands you see executing collabs flawlessly are running a parallel approvals workflow that nobody outside the deal ever sees.

And then there is the morals clause, which is the marketing clause’s evil twin. The Adidas-Ye deal had one (nowadays, almost all deals involving celebrities have them, and most involving brands have them, too). It gave Adidas the right to terminate their contract for reasons including a felony conviction, bankruptcy, mental health issues, or any actions that could bring disrepute, contempt, scandal to Ye or tarnish the Adidas brand. Every meaningful celebrity collab signed in the last five years has a morals clause that looks something like that. The negotiation is over what counts as disrepute, who decides, and whether the licensee gets a cure period.

If you are the celebrity, you fight for narrowness and notice. If you are the brand, you fight for breadth and immediacy. The negotiated reality usually lives somewhere in between.

Who Pays for the Product

This is the question creatives forget to ask until the invoices arrive.

Production funding can flow either direction. In a classic licensee-pays structure, the brand that manufactures the product fronts the cost of materials, tooling, and production, and recovers it through sales. In a licensor-pays structure, often used when the licensor has manufacturing capacity and the licensee is essentially renting the brand, the dynamic flips.

The interesting cases are the hybrids. A celebrity collab where the brand fronts production but recoups against the celebrity's royalty stream until break-even. A luxury collab where each side funds its own piece and the parties true up at the end of the season. A capsule where one party funds production and the other funds marketing and the revenue split reflects the asymmetry. The Beyoncé x Adidas Ivy Park deal was a version of this. Beyoncé received around $20 million in annual compensation while Adidas fronted production and carried the inventory risk, and when 2022 sales came in at $40 million against internal projections of $250 million, the asymmetry of the structure is what made the loss land entirely on the Adidas side. The artist gets paid the same whether the line sells through or sits in a warehouse. The brand finds out, on the back end, what its production funding actually bought.

The clause that matters here is the recoupment definition. Recoupable against what, calculated how, audited by whom. A poorly drafted recoupment clause can turn a collab that sold through into a collab that paid the licensor zero, because every dollar of revenue got eaten by an undefined production overage.

Termination and What Survives

Every collab ends. The good ones end on schedule. The interesting ones end early.

The termination clause has to specify the triggers, the cure rights, and the surviving obligations. Triggers are the easy part. Material breach, insolvency, change of control, morals events. Cure rights are the negotiated part. How long do you have to fix the breach before termination kicks in. Surviving obligations are the part that matters most and gets drafted last.

What survives termination, in a well-written collab agreement, includes the indemnities, the confidentiality, the IP ownership, the audit rights, and the sell-through period. What does not survive, ideally, is any continuing license to use the marks. The mark goes home. The product, in whatever form it has reached the market, goes through the wind-down.

The Yeezy unwind, again, is the case study. During a conference call, Adidas CEO Bjørn Gulden confirmed that the dispute had been settled, with neither side owing further payments. That settlement happened years after the original termination, after multiple rounds of arbitration, after public accusations and counter-accusations. Every dollar of that drama traces back to a clause that either was or was not in the 53-page licensing and endorsement agreement signed in May 2016.

The deal you sign at the start is the deal that governs the unwind. We’ve seen a number of deals twirl into a dispute due to a lack of a clear termination clause.

A Few Others Worth Watching

The lawyer’s bookshelf of great collabs is longer than this essay can hold, but a few deserve a mention before I close.

Louis Vuitton x Supreme remains the most architecturally elegant deal of the modern era. A streetwear brand licensing its mark up to a luxury house, single season, no extension, paycheck plus terms, and then a clean walk away. The collab that proved the streetwear-luxury bridge was real.

Adidas Yeezy, in its 2015 to 2020 prime, generated more wealth for an artist than any collab in history. The Boost 350, the 700, the Foam Runner. A creative output that defined a decade of footwear. The lesson is not that the deal failed. The lesson is that a collab is a marriage, and a 53-page licensing and endorsement agreement is a prenup, and the prenup matters most on the day nobody wants to read it.

The collab nobody is talking about yet but should be is Tyrrell Winston x Adidas. Adidas announced a multi-season partnership with the artist, and the first drop is the Lightblaze POD, covered in a stretched and easeled natural tan artist’s canvas fabric with subtle brush strokes of white added throughout, with Winston’s signature embroidered into the lateral heel like an artist’s signature at the bottom of a painting. A $180 sneaker designed, explicitly, to look better as it falls apart. Tyrrell’s whole practice is about discarded objects and embedded history. The deal here is multi-season, which is rare for an artist collab, and the structural bet is that the relationship deepens over time the way Daniel’s did with Tiffany. Which is the right bet. The collabs that compound are the ones nobody mistakes for one-offs.

When I started writing these deals, the word “collab” was not yet a word in the fashion vocabulary. The form was a trademark license, the conventions were borrowed from celebrity endorsement, and the assumption in most rooms was that the artist gave up their IP and went home with a check. Sixteen years later, the collab is no longer an adjacent activity at the edges of a brand’s calendar. It is, for many houses, the calendar. It drives the windows on Fifth Avenue and the queues outside the pop-ups in SoHo and the resale economy that runs underneath both. It is how heritage brands signal cultural relevance and how new brands borrow the institutional weight they have not had time to build. It is, in short, how fashion now talks to the rest of the culture.

The clauses in this essay are the grammar of that conversation. They are also, increasingly, the place where the most interesting questions in fashion law are being worked out: what does it mean to license a mark that lives as much in the resale market as in the primary one, what does quality control look like when the brand and the artist are inventing the product together, what does termination mean for goods whose cultural value may outlast the legal partnership that produced them. These are not settled questions. They are being settled, deal by deal, in the form files that lawyers in offices in New York and along Avenue Montaigne and Via Montenapoleone are revising in real time.

We have been in those rooms for a lot of it, and we intend to be in them for the next chapter, whatever and wherever those chapters roam! .

The partnerships that last are the ones where both parties treat the document as a beginning, not a closing. The ones where the brand team is willing to say yes when the easier answer is no. The ones where the artist or smaller brand earns the right to push the house somewhere it would not have gone alone.

The license is the bridge between the collaborators.

Sometime last year I looked around at the work we were doing and realized something.

The firm was thriving. The client roster was bigger and more impressive than it had ever been. Litigation was running, transactions were closing, the trademark portfolio work was deeper than ever. By every conventional metric, things were good.

But the artists + creatives were quiet.

The creators, founders, emerging fashion designers and others who had been the gravitational center for us in its early years — those clients were a smaller share of our practice than they’d been at any point I could remember. Not because we’d turned them away. Because the work for brands, companies, and others had grown really fast over a 3-year period.

Without question, working with these kinds of clients is some of the most interesting work we do. The problems are intricate, the stakes are real, the clients are sharp, and we are, candidly, passionate about it. There is a particular kind of craft involved in advising a hundred-year-old brand on a question it has never faced before, or steering a global company through a litigation that could reshape its entire category. That work is part of who we are and we love it!

But there is a difference between a firm that also serves artists and a firm whose center of gravity is artists. I was nervous of drifting away from that spiritual center, not just because I love artists, but for two other, real reasons.

First, we deeply believe that the work we do for early stage artists and creatives make us better lawyers because artists are the original entrepreneurs, able to nimbly craft solutions for age old problems, often times with highly limited resources. And second, our clients — even or especially the big brands — select us because of who we are and the work we do for the creative class.

So we made some changes. These changes were small, but intentional and meaningful. And in the course of making them, we ended up with a clearer picture of what the firm actually is than I’d had at any point in the previous fifteen years.

That picture is a three-layer cake.

The bottom layer: the engine.

This is the work that drives the business. Litigation, transactions, full-service IP for global brands and operating companies. Trademark portfolios across dozens of jurisdictions. M&A diligence. Privacy and trade secret work. Federal court litigation in IP, commercial, and employment matters.

The clients here are sophisticated, demanding, and operate on timelines and budgets that require institutional capacity. The work is intricate, consequential, and genuinely fascinating. There is, importantly, a lot of it.

This layer is the broadest layer of the cake because it has to be. It’s what makes the firm a firm and not a passion project. But it’s also broad because we’re good at this work and we want to do more of it. The two things are not in tension.

What’s true is that the engine isn’t all of who we are. And for a long time we hadn’t articulated what the rest was.

The middle layer: scale.

This is the layer most law firms don’t have, because most law firms haven’t built one.

We have, over the last several years, developed our own legal technology products. They aren’t accessories. They are part of the practice itself, and increasingly part of how we deliver value at scale to the engine clients on the bottom layer and to the artists and creatives on the top.

The first product is Thanks Mom™. It’s an AI-driven trademark search and clearance tool. The simplest way to describe it is this: it does the work of the most expensive trademark search products on the market, at a fraction of the cost, and in many respects does it better. The economics matter. Trademark clearance is one of the most consequential early decisions a brand makes, and historically it has been priced like a luxury good. We didn’t think it should be. So we built something that wasn’t. Eventually, we see Thanks Mom handling most of the work any trademark department would have to do, whether in house or in firm. It’s affordable, easy to use, remarkably accurate, and well-designed.

The second product is Take Care™. It’s a legal technology platform that continuously monitors the terms and conditions of enterprise software and vendor agreements, flags changes in real time, and translates them into clear, practical risk insights for legal, compliance, and business teams. Instead of relying on static contracts that quietly evolve, Take Care creates a live system of record across tools, compares terms across vendors, and routes meaningful changes to lawyers for quick analysis and response — helping companies stay ahead of shifting obligations while reducing cost and friction. And the use case extends well past software contracts. Large companies can use Take Care to manage dynamic legal relationships across entire networks of counterparties: licensees, tenants, franchisees, vendors, marketplaces, users, and other distributed ecosystems where terms, policies, and compliance obligations are constantly changing.

Together, these tools do something I wasn’t entirely sure was possible when we started building them. They let us serve more clients, more deeply, without scaling headcount in the linear way most firms have to. They let an emerging brand access the kind of trademark clearance work that used to require a six-figure budget. They let an in-house team monitor obligations at a scale that would otherwise require an army of associates. They are, in a literal sense, capability.

This is the layer that lets the firm be different sizes for different clients without losing the throughline of who we are.

The top layer: the soul.

This is the smallest layer of the cake. It is also the most sacred.

The top layer is the artists and creatives. The emerging fashion designers building their first collection. The musicians figuring out publishing for the first time. The visual artists navigating their first significant collaboration or their first significant dispute. The founders whose entire company exists because of an idea and a few thousand dollars and a willingness to risk both.

Some of them will become global brands, and some of them will not, and the firm’s relationship with them is not contingent on which one they end up as. It is contingent on something more important than that.

We made two specific commitments to make sure this layer didn’t get squeezed by the natural gravitational pull of the engine work.

The first is the Creative Legal Alliance, or CLA. It’s a year-long, pro bono legal education program for ten emerging fashion brands a year. They get real legal support across the year — not a one-off consultation, not a webinar, but ongoing counsel through the actual decisions and inflection points that shape an early brand’s trajectory. We chose fashion specifically because it sits at the crossroads of so many of the disciplines we care about: IP, branding, identity, distribution, culture. And we chose ten because that’s the number where we can actually serve people well, rather than create the illusion of help at scale.

The second is a standing commitment to discounted rates for emerging artists and a defined number of pro bono representations of artists every year. Not a one-time gesture. A line item. Built into how the firm operates, the way you’d build any other recurring obligation into a budget. The point is that it isn’t supposed to feel like generosity. It’s supposed to feel like infrastructure.

How the layers feed each other.

Here is the part that took me the longest to see, and that I now think is the actual reason the cake works.

The layers are not separate practices that happen to share an office. They are nutrient systems for one another, and the nutrients run in both directions.

The big brands at the bottom of the cake do not, as a rule, want to feel like big brands. The good ones — the ones we like working with — want to feel nimble, culturally fluent, capable of moving the way an emerging brand moves. They are constantly trying to recapture some part of the energy they had when they were small. They watch what the artists are doing, what the streetwear founders are doing, what the cultural-IP edge cases are doing, because that’s where the next decade of branding and consumer behavior is being invented. When we sit across the table from them, our credibility is not just our technical chops. It’s that we are demonstrably in conversation with the people inventing what’s next, every single day. The top-layer work is what makes us useful to the bottom-layer clients in ways a pure corporate firm can’t replicate.

The reverse is also true. The artists and emerging brands at the top of the cake do not want to be advised by a small-time lawyer who can only think small. They want to be advised by someone who has stood in front of a federal judge defending a billion-dollar mark, who has structured a nine-figure transaction, who has navigated a global trademark portfolio. They want a lawyer whose default frame is “this is going to scale” rather than “let’s see what happens.” The bottom-layer work is what gives us the institutional muscle to be genuinely useful to the top-layer clients on day one, before they’re big enough to need it.

Each layer makes the others better. Take any of them away and the rest weaken.

A firm without the top layer drifts. A firm without the engine collapses. A firm without the middle layer is, in 2026, a firm that has decided to compete with one hand tied behind its back.

Why a cake.

The metaphor isn’t decoration. It’s load bearing.

Most service businesses, including most law firms, have a single layer. They have a client list, and they have services they sell to that list, and the math works or it doesn’t. The shape of the business is flat.

A cake is not flat. The bottom layer holds up the middle layer, which holds up the top layer. Each one depends on the ones below it to exist at all. You cannot have the top layer without the foundation, and the foundation by itself isn’t a cake — it’s just a slab. The structure is the point.

What it actually means.

What it actually means is this. The cake is a 360-view of who Jayaram is right now.

I wrote recently about Grant Achatz’s essay on Alinea, and the line that stayed with me was his warning that any creative venture eventually has to ask whether it’s becoming a museum of itself. That question applies to law firms as much as it applies to restaurants. The cake is my answer to it. It’s not a static picture of what we used to be, or what we aspire to be. It’s a picture of what we are today.

We are a fierce advocate for some of the biggest and most interesting brands in the world.

We are technology partners to companies seeking to automate parts of their legal departments.

And we are, as we have always been, an ally to the global arts and fashion communities, because we Enable Original Ideas®. That’s our mission, and it has been from the beginning.

After all, we’re Lawyers for Innovators®.

— Vivek

There is a version of a law firm that everyone in the industry understands.

It is structured, hierarchical, and predictable. It values precision, precedent, and stability. It trains lawyers to analyze from a distance, to observe rather than participate, to interpret rather than immerse. It is built to preserve knowledge, to manage risk, and to deliver answers that are correct.

That model works. It has worked for a very long time.

But it never quite felt right to me.

Not because it was wrong, but because it didn’t seem designed for the people I wanted to work with, or for the kind of work I wanted to do.

And so, from the beginning, I approached the problem differently. I wasn’t trying to build a better version of an existing law firm. I was trying to build something that didn’t really exist in the category at all.

A brand.

Building a Brand in a Category That Doesn’t Have One

Law is a trillion-dollar industry with almost no recognizable brands.

There are well-known firms, of course. Names that carry weight within certain circles. But those names don’t operate the way brands do in other industries. You don’t feel them. You don’t associate them with a point of view, a cultural position, or an identity that extends beyond the work itself.

I wanted to build something different.

Not just a firm people could hire, but a firm people could understand—intuitively. Something that had a presence, a tone, a point of view that was legible before you ever picked up the phone.

That decision shaped everything that followed.

Model the Company After the Client

Most firms, when they think about growth, look at other firms.

We didn’t.

We looked at the people we were representing.

We studied brands like Louis Vuitton, Supreme, Sub-Pop Records, Margiela, Daft Punk, Apple, and Netflix, not legacy partnerships. We paid attention to SoundCloud, to emerging brands, to the way creative companies build identity and distribute themselves into the world. We paid attention to how culture moves, how ideas spread, how communities form, how attention is captured and held.

Because the clients we were drawn to were not thinking like law firms.

They were thinking like builders.

And if you want to advise builders well, you need to build in a way that feels familiar to them.

When your identity mirrors your clients’, trust is not something you have to manufacture. It is immediate.

Also, who doesn’t want to consider themselves an innovator? Hence, Lawyers for Innovators® is a pretty welcomed tagline!

Perception Is Strategic

One of the early lessons was that perception is not just cosmetic.

We made a decision, consciously, to look like something larger than we were. Not in a deceptive way, but in a directional way. We invested in design, in how we presented ourselves, in how the firm felt from the outside.

We looked like a $20 million a year firm before we were anything close to that.

That changed the rooms we were invited into.

It changed the conversations we were part of. It changed how clients understood what we could do, and what role we could play in their business.

In most professional services environments, perception is treated as secondary to capability.

In reality, it shapes where your capability is allowed to show up.

Manifest and execute.

The Advantage of Not Having History

Large law firms have something incredibly valuable: institutional memory. Decades of experience, embedded knowledge, systems that have been refined over time.

They also have something that is harder to see from the outside.

Inertia.

When a firm has been operating for fifty or seventy-five years, it develops a set of expectations about how things are done. Decision-making becomes distributed. Risk tolerance narrows. The ability to experiment and to try something that doesn’t fit the model becomes constrained.

We didn’t and don’t have that.

We didn’t have to reconcile new ideas with old structures. We didn’t have to convince a large partnership to move in a different direction. We didn’t have to optimize for extracting value from a model that was built in a different era.

We could build forward.

That freedom is easy to underestimate. It is also one of the reasons we’ve been able to do things that would be difficult in a more traditional structure.

Cultural Fluency as a Legal Skill

Over time, it became clear that the most valuable thing we could offer clients was not just top-tier legal knowledge (which in our view has always been table stakes).

It was context.

Understanding how people interpret brands, how signals move through culture, how something is going to be received—not just legally, but socially—turns out to be a critical input in almost every decision we help clients make.

That is particularly true in areas like trademark law, where the question is often not just whether something is legally permissible, but whether it creates the right impression in the mind of a consumer.

Most lawyers are not trained to think this way.

They are trained to analyze rules.

We are trained to analyze behavior.

That difference has become a competitive advantage.

Simplicity, Clarity, and the Work Itself

Another principle that has shaped how we operate is the idea that simplicity is not the absence of complexity.

It is the result of understanding it deeply enough to remove what doesn’t matter.

We built the practice around translation—taking legal ideas that are often dense, abstract, and difficult to apply, and making them usable for the people who need to make decisions quickly.

That requires discipline.

It also requires clarity.

For me, that clarity came, in part, from subtraction. Sobriety changed how I think and how I operate. It removed noise. It forced a kind of precision—not just in communication, but in judgment.

Better decisions tend to come from cleaner thinking.

And cleaner thinking often comes from removing what isn’t essential.

The Way You Work Is the Product

In most law firms, the work is the product.

In our case, how we work is also part of it.

The design of our spaces. The way we run meetings. The people we hire. The expectations we set around how we operate—these things shape not just the internal culture, but the kind of work that comes in the door.

We built a firm of runners. People who sleep. People who take care of themselves.

That sounds unrelated to legal performance.

It isn’t.

Energy, focus, and clarity are competitive advantages in a profession that often ignores them.

Personal Brand as Distribution

The legal industry has historically been built on institutional reputation.

We took a different approach.

We built a social platform—not as an end in itself, but as a way of distributing how I think. So that by the time someone needed a lawyer, they already understood the perspective we would bring.

In an environment where most firms rely on referrals and closed networks, that kind of visibility changes the dynamic.

It allows you to reach people before they enter the market.

And it allows the market to understand you on your own terms.

The Creative Life Is the Work

There is a tendency to separate creative life from professional life.

To treat music, art, and other forms of expression as something adjacent to “serious” work.

That separation never made sense to me.

Being a musician, running a record label, engaging with creative communities—those were not distractions from building a law firm.

They were the foundation.

They shaped how I understood risk. How I thought about ownership. How I saw the relationship between culture and business.

I didn’t build this firm despite those experiences.

I built it because of them.

The Bridge

If you look at the legal industry right now, you can see two clear poles.

On one side, legacy institutions that are credible, experienced, but often slow to adapt.

On the other, pure technological disruption—fast, efficient, but not yet fully trusted for high-stakes decisions.

The future does not belong entirely to either.

It belongs to the bridge.

Firms that can operate with credibility and adaptability at the same time. Firms that understand both how law has been practiced and how it needs to evolve.

That is the space we have been trying to occupy.

The Three-Layer Cake

The way I think about the firm now is as a three-layer structure.

At the base is the engine: high-value legal work. Litigation, transactions, the core services that drive revenue and define capability.

Above that is scale: technology that extends that work, makes it more efficient, and allows us to operate at a different level.

At the top is the soul: the artists, founders, and emerging creatives who are often the most interesting, the most experimental, and, in many ways, the most aligned with why we built this in the first place.

It is the smallest layer and it is also the most important.

What We Actually Built

Looking back, I don’t think we set out to build a different kind of law firm.

We set out to build something that felt true to the people we wanted to work with.

A firm that understood companies because it behaved like one.

A firm that understood culture because it participated in it.

A firm that could operate in the present, rather than optimizing for the past.

Everything else followed from that.

And in a category that historically hasn’t had brands, that difference has started to matter.

— Vivek

Likelihood of Confusion

It is the heart of the Lanham Act, and it’s the standard a court, the TTAB, or a USPTO examiner uses to decide whether one mark infringes another. You have committed trademark infringement when your use of a mark is likely to confuse consumers into believing your goods or services are affiliated with, sponsored by, or sourced from someone else.

Every circuit has its own list of factors for answering that question. The Second Circuit calls them the Polaroid factors (based on a case involving Polaroid). The Ninth Circuit calls them Sleekcraft (also based on a case). The Federal Circuit and the USPTO use the thirteen DuPont factors. The First, Third, Fifth, Sixth, and every other circuit have their own variations.

They differ at the margins, but they all converge on roughly the same inputs: the strength of the senior mark, the similarity of the marks, the proximity of the goods and services, the likelihood the senior user will bridge the gap, evidence of actual confusion, the junior user’s intent, the quality of the junior user’s product, and the sophistication of the relevant consumers.

If you ask most trademark lawyers to walk through these factors, they will. Capably. And then they will write you a memo.

That is the problem. Most of these memos never actually account for how a consumer experiences the purchase.

The Two Extremes Are Easy

At one end of the spectrum sits the counterfeit. Someone is using your mark in essentially the same way on essentially the same product, and the entire purpose is to dupe a consumer into buying the fake instead of the real. There is no analytical work to do. The factors all break the same direction because the defendant designed them to.

At the other end sits the case where two marks may even be identical and there is still no plausible confusion. One sells sneakers, the other sells enterprise software. Or one is a $25 shirt and the other is a $250,000 yacht. The products live in such different parts of the consumer’s life that no one is confusing them at the point of purchase.

Almost no real cases live at either end.

Almost all of them live in the middle.

And the middle is where the doctrine stops being as useful, and the consumer takes over.

Commercial Impression Is a Consumer Question

Trademark lawyers like to talk about “commercial impression.”

It is a clean phrase that sounds right in a brief.

What it actually requires is much harder: standing where the consumer stands, in the moment of the purchase decision, and asking what that person actually thinks.

That is not just a legal question. It is a question of culture, behavior, and context.

Most likelihood of confusion analysis collapses into two shortcuts. Are the marks similar? Are the goods related? Yes and yes, there is a problem. No and no, there isn’t.

Both shortcuts produce bad work.

The interesting cases turn on the factors lawyers underweight.

Degree of Care Does the Work

The most underweighted factor, in my experience, is the degree of care that consumers actually exercise in the relevant market.

Price is the obvious lever. We have won jury trials on it. When a product or service costs hundreds of thousands or millions of dollars, the buyer is not browsing. The buyer is exercising a level of care that no visual similarity is going to override.

But price is only part of it.

In the modern marketplace, the more interesting work is structural.

We had a case where the only way to buy our client’s product was to be admitted into a gated community and issued login credentials. There is no plausible confusion in that posture. The consumer is not encountering the brand casually. The consumer is going through a vetted process before a transaction can even occur. Care is built into the architecture of the sale.

The Fortanix Trial

The case I think about most in this regard is Fortanix.

We defended Fortanix in a three-week jury trial in the Northern District of California against Fortinet. Fortinet alleged that the use of a similar name in the cybersecurity space was likely to confuse customers. On paper, the case looked difficult. The marks were similar. The product categories overlapped.

But the case was not difficult once you understood the buyer.

We cross-examined the plaintiff’s sales and marketing teams and walked the jury through what an actual enterprise cybersecurity sale looks like. Sales cycles measured in months. Executive-level approvals. Multiple layers of technical review. Licensing fees in the six and seven figures.

The buyer of an enterprise security platform is not a consumer in any meaningful sense. It is a procurement process.

The idea that such a buyer would confuse one company for another because the names share a root collapsed once the jury saw how the product is actually purchased.

The jury found no infringement.

The doctrine did not win that case alone.

The buyer did.

Adidas v. Thom Browne

Another case that illustrates this better than most treatises is Adidas v. Thom Browne.

Adidas sought millions in damages and argued that Thom Browne’s four-bar design infringed its three-stripe mark. Adidas presented survey evidence suggesting that a significant portion of respondents associated the Thom Browne design with Adidas.

On paper, that is the kind of evidence that wins cases.

It did not win this one.

After trial, the jury returned a verdict for Thom Browne in roughly two hours.

Why?

Because the buyer mattered.

Thom Browne argued that its customer—a luxury consumer spending significant money on fashion—was not going to confuse its product with Adidas. The brands occupied different positions in the market, and the purchasing process reflected that.

When you are buying Thom Browne, you are not accidentally buying sportswear.

The degree of care, combined with brand positioning, did the work that visual similarity could not undo.

The Work Happens Outside the Office

This is where most lawyers fail.

The work does not happen in the memo.

The work happens outside the office. And most lawyers are not trained to go there.

If your client makes a consumer product, you need to walk the aisle. See where it sits. See what it sits next to. Understand how a consumer encounters it. If your client sells sneakers, you need to be in the stores, on the apps, in the resale markets where the buyer actually lives. If your client sells enterprise software, you need to sit through demos and understand what a six-month procurement cycle feels like. If your client is a fashion brand, you need to be at the shows, in the showrooms, in the environments where the brand is actually experienced.

This is not marketing, it’s field research for the case you may have to try.

The Real Case

The doctrine gives you the questions.

The consumer gives you the answer.

Most of our wins in this area come from the same place. We did the work to understand who the buyer actually is, what care they actually exercise, and how the product is actually encountered. Then we showed the jury what we saw.

That is the case.

Not the factors, and not the framework, but the buyer.