You Belong with IP: Can Taylor Swift Really Trademark Her Voice to Fight AI Deepfakes?

Last week, Taylor Swift’s company, TAS Rights Management, filed three new U.S. trademark applications: two for short audio clips of her speaking and one for a concert image showing her onstage in a sequined bodysuit with a pink guitar. Commentators immediately connected these filings to the explosion of AI-generated “Taylor Swift” deepfakes such as fake endorsements, fabricated political statements, and explicit images circulating online.

The filings spark a bigger question for artists and brands: Can trademark law be used to protect your voice and likeness from AI misuse?

1. Do These Trademarks Directly Stop AI Deepfakes?

In a word, no. Trademarks do not directly regulate AI or “ban” deepfakes. Trademark law is still about consumer confusion and source identification, not a general right to control all uses of your image or voice. To the extent that trademarks can protect against unauthorized AI uses, the AI uses themselves would have to constitute trademark infringement.

What Swift filed are classic trademark applications: two “sound marks” consisting of spoken intros promoting her album “The Life of a Showgirl” on Amazon Music and Spotify, and one “design mark” covering a distinctive live performance image. To be registered, these marks must function as brand identifiers for specific goods and services, just like any other word, logo, or jingle.

That means:

• The registrations, if granted, would target infringing uses that create confusion about sponsorship, endorsement, or origin—such as an ad or product using a confusingly similar “Hey, it’s Taylor Swift” voice clip or lookalike concert image to sell something.

• They would not reach every AI deepfake, especially purely noncommercial content or situations that constitute fair use such as descriptive fair use (defendant uses the mark in a descriptive sense rather than as a source identifier for defendants’ own goods or services) and nominative fair use (defendant uses the mark to refer to plaintiff’s own goods and services and does nothing to suggested sponsorship, affiliation, or endorsement by the plaintiff)

While AI concerns almost certainly motivated the filings, commentators have framed them as a “novel legal strategy” to address an AI protection gap. The applications themselves plug into traditional trademark doctrines rather than any new AI-specific right. It is reasonable to suspect Swift’s team is building extra tools for future enforcement against AI-assisted impersonation, not that these filings, by themselves, “solve” deepfakes.

2. Why Bother With Trademarks if You Already Have Right of Publicity?

Celebrities already have a powerful weapon against unauthorized commercial exploitation of their identity: the right of publicity. But that right is primarily a creature of state law, and that creates pains in an AI-driven, internet scale world.

Here are some reasons why federal trademark registration is attractive on top of publicity rights:

• Nationwide, uniform protection. Right of publicity is governed state-by-state, with different scopes, exceptions, and choice of law issues. A federal registration under the Lanham Act gives Swift a single, nationwide right to stop confusing uses of the registered sound and image in commerce, without relitigating state law nuances in every case.

• Powerful Lanham Act remedies. Trademark plaintiffs can access anti-counterfeiting remedies, including statutory damages for use of counterfeit marks on goods, seizure and destruction of counterfeit goods, treble damages and enhanced monetary awards in appropriate cases, and injunctive relief tailored to online marketplaces and platforms. They may also recover attorney’s fees in “exceptional” cases, which is helpful when dealing with willful copycats or organized counterfeiters.

• Counterfeit merch and brand lookalikes. Many of Swift’s real world problems are not only AI voice clones but also a flood of unauthorized merchandise and advertising that trades on her recognizable image and catchphrases. A registered image mark capturing her distinctive stage pose and outfit gives her an additional hook to go after sellers of lookalike posters, apparel, or packaging that reproduce or closely imitate that visual.

• Platform-acing leverage. Platforms and marketplaces are used to triaging trademark complaints and have established takedown pipelines for registered marks and counterfeits. A registered sound or image mark may make it easier to obtain rapid removal of impersonating ads and products than a more fact intensive right of publicity claim.

Put differently, Swift’s move is likely as much about counterfeit merch and unauthorized endorsements as it is about AI deepfakes. Trademark law supplies clear, codified rules and remedies that fit neatly into platform enforcement systems, whereas state law publicity claims can be slower and less predictable.

3. Is She Likely to Get These Trademarks?

Swift is not the first celebrity to test this strategy. Matthew McConaughey has already secured multiple U.S. trademark registrations covering his voice, catchphrases (including his famous “alright, alright, alright”), and short video clips, explicitly to combat AI-driven impersonations and unauthorized endorsements. That history suggests the U.S. Patent and Trademark Office (USPTO) is open, at least in principle, to treating certain spoken phrases and likenesses as registrable marks.

But the path is not guaranteed. From the USPTO’s perspective, several issues will matter:

• Does the sound function as a “source identifier”? The Trademark Manual of Examining Procedure makes clear that sound marks must “assume a definitive shape or arrangement” and create an association in the hearer’s mind between the sound and a particular source of goods or services. Generic or purely informational speech can be refused as not functioning as a mark.

• Is there evidence of use in commerce? Applicants for sound marks must show that the sound is used to brand goods or services. If Swift’s clips are already being used consistently in her official Amazon/Spotify promotions, that bolsters her case; if not, the examining attorney may push for more evidence or a different basis of filing.

• Is the image distinctive as trade dress or a design mark? Stylized images of a performer can be registered if they are used consistently to identify the source—think of a particular posed photo that appears across official merch and marketing. The more that image has become shorthand for “Taylor Swift” in commerce (for example, as key art for the Eras tour), the stronger her argument that it functions as a mark rather than a mere photograph.

• Are there descriptiveness or failure to function refusals? The examining attorney could take the view that “Hey, it’s Taylor Swift, and you can listen to my new album…” is more of an ordinary advertising phrase than a mark, at least without substantial evidence of acquired distinctiveness. Sound mark practice is still evolving, and celebrity “voice intros” are a relatively new category.

Given McConaughey’s success with a suite of voice and image marks, and the USPTO’s existing framework for sound marks, Swift has a credible shot, especially if TAS Rights Management can show consistent, branded use of these clips and the concert image across her official campaigns. At the same time, registering spoken phrases in advertising is still a novel use of trademark law that has not been tested extensively in courts, so some level of pushback or narrowing from the USPTO is realistic.

4. What This Means for Artists, Labels, and Brands

Swift’s filings highlight a larger shift: high-profile talent is now treating their “digital self” voice, likeness, and catchphrases as brand assets that should be fenced in with the full toolkit of IP and publicity laws, not just copyright.

For creators and companies, a few practical takeaways:

• Start cataloging distinctive audio and visual branding elements you actually use in commerce: recurring intro lines, sonic logos, signature poses or illustrations.

• Consider whether federal trademark registration makes sense for any of those elements, especially if you are already battling knockoffs, spoofed ads, or AI-generated impersonations that look and sound “close enough” to confuse consumers.

• Do not view trademarks as a replacement for right of publicity, privacy, or platform policy remedies. In an AI-first environment, the most effective enforcement strategies combine multiple legal theories and are tailored to the specific platform and use.

Kronenberger Rosenfeld’s advertising, IP, and tech litigation teams are already working with artists, brands, and platforms on these questions, including how to structure enforcement campaigns that account for AI-assisted infringement. Contact us today to discuss your legal needs through our online submission form.