Taylor Swift Files Trademark Applications to Fight AI Voice Clones
Taylor Swift's representatives filed trademark applications for her spoken phrases "Hey, it's Taylor" and "Hey, it's Taylor Swift" as a legal workaround against AI-generated voice clones. The move reflects a growing gap in federal law: the Lanham Act, passed in 1946, was never designed to address deepfakes or synthetic voices.
Swift joins other celebrities like Matthew McConaughey, who filed for his signature phrase "alright, alright alright," in using trademark law as a stopgap protection. Without a federal deepfake statute, brand owners and talent representatives are turning to sound mark registration-a strategy with significant limitations.
How Sound Marks Work Under Trademark Law
A sound mark is a trademark in audio form. Streaming platforms' opening chimes and tones function as sound marks, creating distinctive auditory signatures that identify the source of content. Like visual logos, they must meet three legal requirements.
Distinctiveness: The sound must identify a single source, either through inherent uniqueness or through evidence of secondary meaning-longstanding use, advertising, media coverage, and consumer recognition. A spoken phrase qualifies only if consumers perceive it as identifying one source, not as everyday speech.
Non-functionality: The Lanham Act bars registration of marks essential to a product's use or purpose. A film's promotional tagline is protectable because it promotes the film. Dialogue within the film itself is not, since it serves the product's function rather than identifying its source.
Specimen requirements: Applicants must provide evidence of a direct association between the mark and the services. Swift's applications include recordings where she opens with the target phrase and continues with promotional messaging about a new album or streaming platform.
The Patent Office's Assessment
The US Patent and Trademark Office will determine whether Swift's phrases are inherently distinctive for entertainment services or require secondary meaning evidence. The real vulnerability lies not in initial approval but in post-registration challenges.
A 2023 trademark decision broadened the types of acceptable evidence for sound marks, allowing recordings played in promotional contexts to qualify as valid specimens-even without physical affixation to packaging.
A Narrow Shield Against Deepfakes
Sound mark registration protects only against confusingly similar uses tied to related goods or services. This creates a critical gap in deepfake enforcement: an AI-generated voice mimicking Swift's voice, but not using the exact trademarked phrase, typically falls outside the mark's scope.
A deepfake of Swift endorsing a product might not infringe a mark limited to "Hey, it's Taylor" unless consumers hear substantially the same phrase in a similar context. Right-of-publicity laws-which protect the voice itself rather than a specific phrase-would be more effective, but they vary by state and lack federal consistency.
Congress has proposed legislation including the NO FAKES Act and NO AI FRAUD Act, but neither has passed. Until federal law addresses AI voice cloning directly, sound marks remain incomplete protection.
A Layered Defense Strategy
Sound marks work best as one component of a broader enforcement strategy. Successful registration creates nationwide rights with access to statutory damages and Customs recordation-tools unavailable under state law alone.
Brand owners should combine sound marks with copyright protections, voice likeness restrictions, and contractual safeguards that address unauthorized AI training. The most effective approach registers short, distinctive vocal phrases used consistently across promotions, building evidence of distinctiveness while maintaining a broader portfolio of legal protections.
Without federal deepfake legislation, this patchwork approach remains the practical standard for protecting voices from synthetic replication.
Your membership also unlocks:
