Swift's TAS Rights Management filed three landmark trademark applications on April 24, testing whether sound marks can build a legal wall around a celebrity's AI-replicated identity.
On April 24, 2026, Taylor Swift's company TAS Rights Management submitted three trademark applications to the US Patent and Trademark Office. Two cover sound marks: specific audio clips of Swift saying "Hey, it's Taylor Swift" and "Hey, it's Taylor." The third protects a visual of her performing on the Eras Tour stage in a sequined bodysuit, holding her signature pink guitar. The Guardian confirmed the filings, and intellectual property attorney Josh Gerben, who does not represent Swift, called them a novel application of trademark law that has never been legally tested at scale.
The move follows a similar filing by Matthew McConaughey, whose voice trademark applications were approved earlier this year after he told the Wall Street Journal in January that the industry needed clear ownership norms for the AI era. Where McConaughey blazed the trail quietly, Swift's filing lands with cultural force. Spotify's most-streamed artist of all time, the subject of explicit AI deepfakes that spread globally in early 2024, unauthorized fake political endorsements shared by President Trump on Truth Social, and Meta AI chatbot recreations, she arrives at this legal frontier with documented evidence of the exact harms the filings are designed to prevent.
Copyright has always been the first line of protection for musicians. It covers specific recorded performances and compositions. What AI changes is the gap between mimicry and reproduction. A model trained on Swift's catalog can generate an entirely new song in her voice without copying a single existing recording, leaving copyright with nothing to grab. That legal void is precisely what trademark law targets.
Gerben explained the theory directly: by registering specific spoken phrases tied to her voice, Swift could argue that any AI-generated audio confusingly similar to the registered trademark violates her trademark rights. The key phrase is "confusingly similar," a standard trademark law uses to evaluate consumer confusion. An AI song that sounds like Taylor Swift, released by an account called "Tayler Swifft," could meet that bar. The image trademark adds a parallel layer: protecting a distinctive visual combination, that specific pose, outfit, and instrument, gives grounds to challenge AI-generated images that evoke her likeness without reproducing any protected photograph.
The Independent reported that attorney Kirk Sigmon of Kell Dann sees this as sound marks meeting deepfake law for the first time. It is untested. Courts have not adjudicated whether a celebrity's speaking voice, rather than a musical performance, qualifies as a trademark. The next litigation that cites these applications will set the precedent.
A Legal Playbook Bigger Than One Celebrity
Swift's filings matter far beyond her personal protection because they establish a replicable template. Every major entertainer, broadcaster, and public figure who has faced AI impersonation now has a roadmap. NBC News reported the filings, noting that Swift has filed hundreds of trademarks throughout her career covering lyrics, merchandise, and brand elements, but this marks the first time she has sought sound mark protection. Others will follow, and quickly.
The convergence with pending legislation sharpens the stakes. Representative Ted Lieu's bipartisan deepfake bill, currently advancing through the House, targets criminal penalties for non-consensual synthetic media. Swift's trademark strategy works alongside that legislation, not instead of it. Criminal law addresses bad actors who distribute deepfakes. Trademark law gives IP holders a civil claim against creators, platforms, and advertisers who use AI-generated likenesses commercially without consent. Both are needed because the threat operates on two tracks simultaneously: malicious personal harm and commercial exploitation.
What It Means for AI Content Platforms
For companies building voice cloning, image synthesis, or synthetic media tools, Swift's filings introduce a new legal liability layer that operates independently of copyright. A platform that generates Taylor Swift's voice on demand, even from a user prompt with no existing recording sampled, now faces potential trademark infringement claims if the output is confusingly similar to the registered marks. That changes product design decisions. Age verification, voice filter blocking, and opt-out registries are no longer just ethical considerations; they become legal risk management.
Forbes noted the broader trend: celebrities using trademark law to counter AI represents an emerging legal strategy that IP holders across entertainment, advertising, and media are watching closely. The advertising industry, which has already seen brands exploited by AI-generated fake celebrity endorsements, faces similar liability exposure. Brands that license celebrity likeness for campaigns now need to verify the authenticity of every generated asset. Watch whether Swift's applications clear the USPTO's examination process, as the outcomes will define whether sound marks become the standard protection mechanism for voice identity in the AI era, or whether courts push creators back toward right-of-publicity statutes and new federal legislation to fill the gap.
Also read: Canonical turns Ubuntu into the AI OS with native model management and edge inference • Monroe Capital's Koenig says AI supercharges only narrow software niches • GitHub Copilot moves to token billing in June and developers face a reckoning
