Taylor Swift just opened a new front in the celebrity war against AI copycats. According to The Verge AI, her team at TAS Rights Management filed trademark applications last week covering two spoken phrases, “Hey, it’s Taylor Swift” and “Hey, it’s Taylor,” along with a stage photo of the singer holding a pink guitar in an iridescent bodysuit. The filings include audio clips of Swift promoting her latest album, The Life of a Showgirl.
Swift’s team hasn’t publicly said the move targets AI mimics. But given her history with AI music knockoffs and sexualized deepfakes, the timing tells the story.
Why Trademarks, Not Copyrights
Here’s the legal gap Swift is trying to close: copyright protects a song, not a voice. That’s why labels have struggled to take down AI-generated tracks that sound like real artists. Universal Music Group had to get creative last year, pulling an AI Drake song by citing the Metro Boomin producer tag at the start, not the vocal performance itself.
Trademarks work differently. As IP attorney Josh Gerben told The Verge AI, a trademark on a signature phrase could let Swift challenge not just identical copies but anything “confusingly similar.” The photo trademark could similarly cover AI-generated images that imitate her stage look.
Matthew McConaughey pulled the same play earlier this year, trademarking clips of himself saying “Alright, alright, alright.” Expect more A-listers to follow.
The Catch
Legal experts The Verge AI spoke to are skeptical the filings will sail through. Alexandra Roberts, a law and media professor at Northeastern, said Swift’s audio specimen looks more like a phrase buried inside a longer ad than a true soundmark. The classic soundmark examples, she noted, are things like the NBC chimes or the MGM lion roar, played in isolation as a brand identifier.
If the USPTO issues preliminary refusals, Swift’s team gets another shot with different specimens.
UCLA law professor Xiyin Tang offered a more pragmatic read. Even a shaky trademark can “warn off unsophisticated infringers” by giving the legal team a registration number to wave around. The deterrent value may matter more than whether it would actually hold up in court.
What Artists Can Already Do
Swift isn’t starting from zero. The current toolkit includes:
- Right of publicity laws in several states, covering misuse of name or likeness
- Federal false advertising and endorsement claims
- Existing trademark registrations on her name, which she can already use to sue for infringement when someone creates likelihood of confusion
- Tennessee’s ELVIS Act, the only state law specifically targeting AI-generated voice copycats of artists
YouTube’s deepfake detection tool exists too, but it only handles faces right now, not voices.
Why This Matters
What stands out here is the asymmetry. AI voice cloning got cheap and good in roughly 18 months. The legal framework to fight it is still patchwork, state by state, statute by statute. Artists are bolting trademark filings onto a system that wasn’t built for synthetic media.
For anyone building or deploying voice AI, the signal is clear. Big-name rights holders are pre-positioning legal claims on signature catchphrases, signature looks, and anything else that could anchor a future infringement case. Training data scraped from public performances is one risk. Generating outputs that sound “confusingly similar” to a registered trademark is another, and that’s the lever Swift is reaching for.
For practitioners, three things to track:
- Whether the USPTO accepts these phrase trademarks or pushes back on the soundmark standard
- Whether more states follow Tennessee with AI-specific voice protection laws
- Whether platforms expand deepfake detection beyond faces to cover synthetic voices
The legal murk won’t clear up fast. Until Congress or the courts settle the bigger question, expect more celebrities to file creative trademarks and dare AI companies to test them.
Full reporting available at the original source.