Taylor Swift filed to trademark the very phrase meant to prove it’s her.
A court already ruled that’s not what stops a clone.
In April, Taylor Swift’s company filed to trademark the very phrase she’d use to prove a call is really her. Nine months earlier, a federal court had already ruled that a voice announcing its own name creates no legal confusion to protect against; that’s the exact mechanism her filing depends on.
Trademark can’t stop a deepfake that announces its own name. In April, TAS Rights Management — Taylor Swift’s company — filed as if it can: three trademark applications with the U.S. Patent and Trademark Office, two of them sound marks for the spoken phrases “Hey, it’s Taylor Swift” and “Hey, it’s Taylor.” The stated purpose: defense against the explicit, non-consensual deepfakes that circulated of her in January 2024. Nine months earlier, a federal court in New York had ruled on the very question her filing turns on, in a different case, for a different voice. That ruling gutted the theory before she ever filed.
That mismatch already played out for someone else. Matthew McConaughey got there first. Eight sound-mark registrations, applied for in December 2023, approved this past December: “alright, alright, alright,” a twelve-second stretch of dialogue, two video clips. His marks are live; Swift’s are still pending examination.
But approval was never really the question. In July 2025, Lehrman v. Lovo asked a federal court whether an AI company cloning a voice actor’s voice infringed his trademark rights. The court said no. Trademark protects against confusion about where something comes from. A deepfake that opens with “Hey, it’s Taylor Swift” doesn’t confuse anyone about the source — it announces it, correctly, as the first line of the con.
What a registered mark actually buys you
Secondary meaning is the whole test. A mark only works once consumers hear it and think of one source, which the Lanham Act calls secondary meaning. A voice that announces itself never gets the chance to earn that: the listener already knows who’s talking before the mark can do its job.
Standing, not a shield. What McConaughey’s lawyers built with eight registrations is a federal cause of action. A mark gets you into court the moment someone infringes it. It doesn’t stop the infringement from happening first.
Approval was never the point. Swift’s applications are pending, McConaughey’s are granted, and it changes almost nothing about whether either one can prevent the next clone. Registration is proof after the fact, not a lock on the door.
The actual fix is still being drafted
The real protection isn’t trademark law at all: it’s a right that doesn’t fully exist yet. The NO FAKES Act, reintroduced in the Senate this June with backing from Google, OpenAI, and the major labels, would create the first federal voice-and-likeness right. Built for this problem, not repurposed from a doctrine meant to protect soup cans and soda logos. Until it passes, the strongest tool on the table is state publicity law, the one claim Lehrman v. Lovo left standing even while it threw out the trademark claim.
This is the same seam Identity Architecture™ is built for, one courtroom over. The question was never whether Taylor Swift is famous enough to be protected. It’s whether the systems now deciding what’s real about her — or anyone who speaks for a living — can verify who’s talking before a court has to. A trademark is a claim filed after the fact. Identity Architecture™ is the machine-readable structure that lets an inference system tell the difference before anyone has to sue.
Seventeen editions in, the pattern holds: the fastest-moving protections are the ones built for a world that stopped existing three product cycles ago. Trademark law was never built to certify a voice. Something will be. Until it is: standing in court is one kind of protected. Being recognized before the room decides for you is another.
TAS Rights Management filed three U.S. trademark applications in April 2026 covering spoken phrases and a visual mark tied to AI-deepfake protection — reported by Variety.
Matthew McConaughey’s team secured eight sound-mark and video-clip trademark registrations, applied for December 2023 and approved December 2025 — reported by NPR.
In Lehrman v. Lovo (S.D.N.Y., July 2025), a federal court dismissed a voice actor’s trademark and copyright claims against an AI voice-cloning company, while leaving state publicity-law claims standing — summarized by Fred Law.