Trademark Law in the AI Era: Why Brands Should Be Filing Like Taylor Swift and Jimmy Kimmel
Trademark law wasn't designed for an AI world. But it may become one of the most effective tools to fight it. Celebrities are already leading the charge — and the same playbook applies to every brand with a mascot, spokesperson, package, or product design worth protecting.
Trademark law wasn't designed for an AI world. The Lanham Act predates the modern internet by half a century, and its drafters were thinking about confusingly similar product names — not synthetic voices, deepfakes, and generative video. But despite that mismatch, trademark law may be turning into one of the most effective tools rights holders have to fight AI-generated content. And the most interesting part is that celebrities saw it before in-house counsel did.
Last week, Taylor Swift filed trademark applications to protect her voice and image. This week, it's Jimmy Kimmel. Earlier this year, Matthew McConaughey did the same. The press is treating it as a celebrity story — voice clones, deepfakes, "Hey, it's Taylor." But that framing buries the lead. There is a real takeaway here for in-house counsel and brand owners, and it has nothing to do with pop stars.
Taylor and Jimmy's playbook applies to every brand. If AI can replicate it, you should be protecting it.
What the Filings Actually Cover
The mechanics of these celebrity filings matter because they reveal the strategy. These aren't ordinary word-mark applications. They're sound marks and design marks deployed deliberately to address an AI threat that copyright cannot fully reach.
- Sound mark: her voice saying "Hey, it's Taylor Swift"
- Sound mark: her voice saying "Hey, it's Taylor"
- Design mark: a specific stage image — pink guitar, multicolored bodysuit, silver boots, pink stage
- Sound mark: his voice saying "Hi, I'm Jimmy… I'm the host of the show… thank you, thank you… thanks for coming… thanks for watching at home"
- Design mark: Kimmel seated behind a desk with his name on a placard
- Design mark: Kimmel standing on the set of his nightly program
- Word/sound marks: his iconic catchphrase "Alright, alright, alright" and other voice/likeness elements
Sound marks aren't new. The USPTO has long registered iconic audio cues — MGM's lion roar, NBC's chimes, Netflix's "tu-dum," the Pillsbury Doughboy's giggle. What's new is using sound marks (and tightly defined design marks) as a deliberate countermeasure against AI voice cloning and image synthesis. As one trademark attorney put it, attempting to register a celebrity's spoken voice for this purpose is a use of trademark registration that hasn't been tested in court before.
Why Trademark — Not Copyright — Is the Right Tool
The reason this strategy is gathering momentum is that copyright has a structural gap when it comes to AI. Copyright protects against direct copying. If an AI system copies a recording, you have a copyright claim. But generative AI doesn't usually copy. It generates new outputs that sound like or look like the original — and that distinction is exactly where copyright runs out.
Trademark works differently. Trademark protects against confusingly similar uses. If a consumer might think your brand is associated with the AI-generated content — even if no recording was copied, even if the output is technically novel — you may have a trademark claim. That broader scope is what makes trademarks particularly valuable in the AI era.
Copyright
- StandardSubstantial similarity to a fixed work
- Best AgainstDirect reproduction or close derivatives
- AI WeaknessGenerative outputs that don't copy any specific work
- RemedyDamages, injunction, statutory damages if registered
Trademark
- StandardLikelihood of consumer confusion as to source
- Best AgainstImitations and look-alikes that suggest association
- AI StrengthReaches "confusingly similar" AI outputs even without copying
- RemedyInjunction, damages, profits, takedown leverage with platforms
This is the structural insight driving the celebrity filings. Copyright handles the recording. Trademark handles the imitation. AI sits squarely in the imitation zone, which is why the trademark side of the toolkit is suddenly the more useful one.
The Real Takeaway: This Isn't Just a Celebrity Story
If you're in-house counsel at a consumer brand, the celebrity headlines are easy to dismiss as someone else's problem. They shouldn't be. The same risks that drove Taylor Swift and Jimmy Kimmel to file are present in any brand asset that AI can replicate — and AI can replicate almost everything that consumers recognize.
Think about what your brand actually owns at the recognition layer. The voice that says your tagline. The character that appears in your packaging. The spokesperson who has fronted your campaigns for fifteen years. The shape of your bottle. The pattern on your box. Every one of those assets is exposed in a generative-AI environment, and every one of those assets is potentially registrable.
Brand Assets to Audit for Trademark Coverage
Mascot Voices
The Geico Gecko. The AFLAC duck. The Pillsbury Doughboy. Distinctive voice performances tied to a specific character.
Mascot Images
Tony the Tiger. The Michelin Man. The Energizer Bunny. Visual representations of brand characters that AI can readily generate.
Spokesperson Voices
That's you, Jake from State Farm. Long-running spokespersons whose voices are part of brand recognition.
Spokesperson Images
The Most Interesting Man in the World. Flo from Progressive. Specific, recurring visual presentations.
Product Packaging
Trade dress — distinctive bottle shapes, color schemes, label layouts, and overall package design that signals source.
Product Design
Distinctive product shapes and configurations whose appearance has become associated with your brand in the marketplace.
If AI can replicate it, you should be protecting it. That's the operative test. Run your brand audit through that lens and the registration priorities reorder themselves quickly.
How Enforcement Actually Works
One reason this strategy is gaining traction is that the enforcement playbook is practical, not theoretical. The first test of these trademarks isn't going to be in federal court. It's going to be a demand letter.
Here's the typical sequence. A trademark registers. Someone — or something — generates content that uses the protected voice or image. The brand's legal team sends a demand letter to the AI platform asserting trademark rights and requesting removal or restriction. If the platform complies, the trademark has done its job without ever being tested by a judge. If the platform pushes back, litigation becomes the next step, but by that point the brand has built a record of demands, refusals, and consumer-facing harm that strengthens its position.
This is exactly how Disney has been operating. In December 2025, Disney sent a cease-and-desist letter to Google alleging that Gemini was being used to generate copies of its trademarked characters. Within roughly a day, Google had pulled down the offending content. No lawsuit, no trial, no precedent-setting opinion — just a demand letter backed by registered marks and a platform that decided compliance was easier than litigation.
Trademark enforcement against AI doesn't require winning at trial. It requires credible threat capacity — registered marks, clean documentation of confusing use, and a willingness to escalate. Most disputes will be resolved at the demand-letter stage. The brands that will win those exchanges are the ones whose trademark portfolios already include the assets AI is most likely to imitate.
What In-House Counsel Should Do This Quarter
Trademark portfolios that were sized for a pre-AI world are almost certainly under-built for the current one. The audit is straightforward. The action steps are too.
Inventory the recognition layer. Walk through every asset that consumers associate with your brand at a sensory level — voices, jingles, characters, distinctive packaging, signature product shapes, recurring imagery. The list is usually longer than the legal team expects, because marketing has spent years building those associations whether or not anyone filed paperwork to protect them.
Identify what's already protected — and what isn't. Word marks and logos are usually filed. Sound marks, character voices, distinctive design marks, and trade dress are usually not. The gap is where exposure lives.
Prioritize by AI replication risk. Not every asset needs the same treatment. Focus first on what generative AI tools can actually produce convincingly today: voices, faces, mascot characters, and visually distinctive packaging. That's where takedown leverage matters most.
Build the file before you need it. Trademark registrations take months. The brands that will be in the strongest position eighteen months from now are the ones filing now, not the ones reacting after a deepfake goes viral. Ensuring your company has a fully loaded file of trademark registrations will provide a critical tool in the fight against AI-generated content.
The Bottom Line
Trademark law wasn't designed for an AI world, but its core mechanism — preventing consumer confusion as to source — turns out to be exactly what's needed when AI can fabricate plausible imitations of almost anything. Taylor Swift and Jimmy Kimmel are testing the strategy at the celebrity level. The same logic applies, with even fewer doctrinal complications, to mascots, spokespersons, packaging, and product design.
The brands that file now will spend the next few years sending effective demand letters. The brands that wait will spend that time watching their identity show up in places they never authorized — without the legal infrastructure to do anything about it.
For deeper guidance on building a trademark portfolio designed for AI-era enforcement, including sound mark and design mark filings, see our Amazo
RSS Feed