TAS Rights Management, LLC filed three USPTO trademark applications for Taylor Swift on April 24, 2026: two spoken greetings and a stage image. They remain applications, but they show how AI imitation is pushing celebrity-brand protection toward trademark filings alongside publicity-rights claims.
Taylor Swift’s Filings Turn A Deepfake Problem Into A Trademark Question
TAS Rights Management is the applicant, and the Associated Press reported the applications were still awaiting examination at the U.S. Patent and Trademark Office. The filings put defined voice and image cues on the record before the next AI dispute arrives.
The sound applications cover the spoken greetings “HEY, IT’S TAYLOR SWIFT” and “HEY, IT’S TAYLOR.” Both are in Class 041 for music information, non-downloadable audio and video, artist news, and related online content. The service descriptions do the legal work: they connect the greetings to places where fans may encounter them as commercial signals.
The image filing is similarly specific. It describes Swift holding a pink guitar with a black strap, wearing a multicolored iridescent bodysuit and silver boots, standing on a pink stage with a multicolored microphone and purple lights behind her. That detail is the point of the filing. The claim centers on a defined performance image fans may connect with her entertainment services.
I read the applications as preparation for an enforcement problem that is already visible. Reuters reported that Swift’s voice and image have already appeared in AI-generated deepfakes, including false ads, fake political endorsements, and explicit images. A trademark filing creates a public record of what Swift’s team says identifies her as the source. If a later advertiser or platform calls a copied sound or image decorative, that record gives Swift a starting point.
Filing early gives Swift’s team that advantage. The relevant sounds and image are already defined when the dispute begins.
Why A Sound Mark Is Not The Same As Owning A Voice
A sound can be a trademark when listeners associate it with one commercial source. NBC’s chimes and Netflix’s opening sound are familiar examples. A human voice is harder because it is identity first, and only sometimes a brand signal.
Swift’s applications sit inside that limit. They claim two short spoken phrases tied to entertainment services. The legal question is whether a distinctive voice can function as a trademark when consumers hear it as a source signal rather than proof that the speaker is famous.
The filings should temper the headlines. A filing starts review, but it does not guarantee registration or a win against an AI platform.
Fame explains why the filings draw attention, but it is not the legal test. The USPTO will look for evidence that those greetings point consumers to Swift’s entertainment services. The office will ask for actual trademark use, not celebrity recognition.
The harder question comes when an AI-generated clip uses a greeting close enough to an applied-for phrase that viewers think Swift approved the message. Copyright law would focus on the recording. Publicity law may address identity. Trademark law asks the narrower commercial question: did the use confuse consumers about source, sponsorship, or approval?
AI creates pressure because the copied sound may not be a pirated recording, yet it can still make the public think Swift approved the clip, the product, or the platform using it. Trademark law is not built for every identity injury, but it is built for that kind of commercial confusion.
Specimens and service descriptions matter because the sound has to operate in commerce, not float as a personal catchphrase. Class 041 gives the phrases a commercial context: entertainment information, audio and video content, artist news, and related services.
If the USPTO accepts the sound marks, Swift would still need facts showing confusion in any later dispute. The registration would be a tool, not a shield that blocks every imitation. For a public figure whose business depends on voice, performance, and trust, though, that tool could be worth having before a fake clip goes viral.
Why Celebrities Are Filing Around Voice And Image Assets
Swift’s filings fit a wider celebrity-brand problem: fake clips travel faster than legal theories. By the time a celebrity responds, the campaign may have already reached the audience it was built to influence or sell to.
The commercial risk is not limited to fans being fooled for a few minutes. A false endorsement can steer traffic, sell products, and attach a celebrity’s goodwill to a message the person never approved.
A celebrity can deny the fake, but the denial rarely travels as far as the clip. By then, the ad may have sold products, harvested attention, or made the next fake easier to believe. A clear filing record has value before the misuse happens.
For celebrities, distribution is now the business problem. AI clips can move through ads, fan accounts, political posts, product pages, and streaming promotions before anyone confirms whether the use is real. A trademark claim becomes stronger when the copied sound or image is not merely imitating a person, but selling, promoting, or pointing audiences back to services the public associates with that person. The channel matters as much as the clip: a fan edit raises different trademark questions from the same cue inside a paid ad, product page, or subscription prompt.
Swift has extensive trademark coverage, from album titles to merchandise and tour filings. The April 2026 applications add a new category to the broader Taylor Swift trademark portfolio: voice and image cues claimed as identifiers for her entertainment services. Reuters connected the move to Matthew McConaughey’s earlier attempt to register voice and image assets tied to his public identity.
The image filing may be the clearest example of the strategy. It points to a particular performance look and asks whether that look has source-identifying meaning. The strongest trademark argument centers on a recognizable commercial presentation.
If a fake ad copies the exact phrase or image, the argument is easier. If the imitation merely sounds like Swift or evokes a stage look without matching the filing, the trademark theory becomes less certain. Those facts will decide the strength of the claim.
What Brand Owners Should Learn From The Swift Filings
For most businesses, the lesson is narrower than the celebrity headline. A brand may include more than its name and logo. If customers treat a sound cue, recurring phrase, package shape, mascot, uniform look, character, or presentation style as pointing to one source, it deserves trademark review.
When I evaluate unusual brand assets, I start with function. Is the asset identifying the business, or decorating the work? A repeated jingle may carry source meaning, while background music usually does not. Product packaging used the same way across a line is stronger than one attractive label. A phrase customers repeat is stronger than copy that changes every campaign. Context, not novelty, decides the answer.
Swift’s filings also show the value of precision. Her team filed for a defined stage image and two defined audio phrases. That same discipline should guide planning a federal trademark filing around unusual brand assets. The more specific the asset, the easier it is to explain what the public is supposed to recognize in a real filing.
Evidence drives the analysis. I look for consistent use, dates, specimens, customer recognition, and a goods-and-services description that matches how the asset appears in commerce. If the asset has only appeared once, or if customers would see it as ornamentation, a trademark application may be premature. If it has become a repeatable brand signal, waiting gives copycats and AI tools more room to blur the connection. The best filing is usually built backward from that evidence rather than forward from a clever idea.
For smaller companies, the evidence may be simpler than celebrity proof: dated website captures, packaging photos, ad records, sales records, and customer references showing the asset doing brand work in the market.
Overclaiming creates its own problem. A business that tries to register every appealing image, phrase, or sound can waste money and invite refusals. The better move is to protect the assets that actually carry brand meaning first, keeping the filing focused on customer recognition instead of wishful thinking. That discipline matters more for assets outside the usual word-mark and logo categories.
Protect Brand Signals Before Imitation Spreads
AI lowers the cost of imitation and accelerates its spread. Swift’s filings show one way a brand owner can move before the test case arrives: identify the signals customers already connect with the source, then create a record while the facts are fresh. Once a fake use gains attention, correcting the market is slower and more expensive than documenting the trademark position early.
In practice, that means separating brand signals from ordinary marketing before imitation blurs the source. I review names, phrases, designs, sounds, product presentation, and filing strategy in that context, then explain what can be protected, what needs more evidence, and where enforcement may become difficult later.
If competitors or AI tools could copy a brand signal your customers already recognize, evaluate it before the confusion starts. You can contact my office to discuss a trademark strategy and decide which assets are worth protecting now.
