In April 2026, Taylor Swift filed a series of trademark applications with the United States Patent and Trademark Office seeking protection over elements of her commercial identity, including short audio phrases such as “Hey, it’s Taylor Swift” and “Hey, it’s Taylor,” along with associated visual branding. Press coverage of the filings has tended to frame the effort as an attempt to “trademark her voice.”

The shorthand description of “trademarking a voice” is imprecise, but it captures something more important: a sophisticated attempt to use trademark law to address a growing gap in legal protection exposed by generative artificial intelligence.1.

Under existing U.S. law, copyright does not protect a person’s voice as such. Copyright attaches to original works of authorship fixed in a tangible medium, which means a particular recording or performance is protected, but the underlying sound or timbre of a voice is not. This distinction, once largely academic, has become central in the era of AI-generated content. Modern generative systems can produce highly convincing vocal imitations without copying any specific recording. As a result, the output may avoid copyright infringement even while clearly invoking a particular artist’s identity.2.

Faced with that gap, Swift’s filings reflect a deliberate pivot toward trademark law, and more specifically toward the doctrine of sound marks. The Lanham Act permits protection not only for traditional word and design marks but also for nontraditional marks, including sounds, so long as they function as indicators of source. The inquiry is not whether the sound is creative in the copyright sense, but whether it is distinctive and has acquired source-identifying significance in commerce. See 15 U.S.C. § 1127; U.S. Patent & Trademark Office (“USPTO”), Trademark Manual of Examining Procedure § 1202.15 (Oct. 2018). By seeking to register short, recognizable audio phrases associated with her public persona, Swift is not claiming ownership of her voice in the abstract. Rather, she is attempting to secure exclusive rights in specific auditory signatures that consumers may associate with her brand.

That distinction is likely to be critical if and when these applications are examined or litigated. A registrable sound mark must do more than exist; it must operate in the marketplace as a mark. The question for the USPTO will be whether the proposed audio clips are used in a manner that signals source, rather than merely conveying content. Courts have long recognized sound marks where a consistent auditory cue identifies a single commercial source, but they have also been careful to deny protection where the alleged mark is merely ornamental or informational.3.

The strategic value of this approach becomes clearer when viewed against the limitations of the right of publicity. Publicity rights have traditionally provided the primary mechanism for celebrities to control the commercial use of their name, image, and likeness. Yet those rights are state-based, vary significantly in scope, and have only recently begun to grapple with AI-generated likenesses. While some jurisdictions have moved toward explicit statutory protection against digital replicas, the law remains uneven and, in many respects, reactive. See, e.g., Midler v. Ford Motor Co., 849 F.2d 460, 463 (9th Cir. 1988) (recognizing misappropriation of a distinctive voice under California law); Waits v. Frito-Lay, Inc., 978 F.2d 1093, 1100–01 (9th Cir. 1992).

Trademark law, by contrast, offers a federal framework with established doctrines such as likelihood of confusion, false endorsement, and dilution, each of which may be adaptable to disputes involving AI-generated imitations. See 15 U.S.C. §§ 1114, 1125(a), (c). If Swift succeeds in registering these sound marks, the practical effect would not be to prohibit all imitations of her voice. Instead, it would provide a basis to challenge uses that are likely to cause confusion as to affiliation, sponsorship, or endorsement. In the AI context, that could include synthetic audio that closely mimics her branded phrases or otherwise suggests that she is associated with a particular product, service, or piece of content.

At the same time, significant doctrinal questions remain unresolved. Trademark law has not yet fully confronted how to assess likelihood of confusion when the allegedly infringing output is generated by an algorithm rather than authored by a human actor. Nor is it clear how courts will treat the role of intermediaries, such as platforms or model developers, in facilitating the creation and dissemination of such content. These issues will require courts to adapt existing frameworks without losing sight of the core purpose of trademark law, which is to prevent consumer confusion and protect the goodwill embodied in a mark.4.

Swift’s filings should therefore be understood less as a claim to ownership of a voice and more as a test case for how far trademark law can be extended to protect identity in the age of artificial intelligence. They illustrate a broader trend in which rights holders are leveraging existing doctrines to fill gaps left by technological change. Until legislatures or higher courts provide clearer guidance on the treatment of AI-generated likenesses and voices, trademark law may serve as a pragmatic, if imperfect, tool for enforcement.

For practitioners and businesses, the implications extend well beyond the music industry. As voice interfaces, synthetic media, and AI-generated content become more pervasive, the concept of brand identity is expanding to include auditory and behavioral cues that were previously difficult to protect. Companies and individuals who rely on distinctive voices or audio branding may find themselves confronting the same questions Swift’s applications raise: whether those elements can function as marks, how they are used in commerce, and what steps are necessary to establish and enforce rights in them?

In that sense, Swift’s filings are not merely a reaction to AI-driven risks; they are an early indication of where trademark law is headed. The next generation of trademark disputes is likely to focus not only on names and logos, but on the broader constellation of signals—visual, auditory, and digital—that consumers associate with a source. The law will have to evolve accordingly, and the outcome of efforts like this one will help define the contours of that evolution.
GCT attorneys Anthony Wenn and Sarah Dunkley routinely deal with intellectual property issues for clients, both individuals and businesses looking to protect their property. They recently gave a presentation on various intellectual property issues to the Technology and Manufacturers Association and can be contacted at GCT.law.

1. See Angela Yang, Taylor Swift Moves to Trademark Her Voice and Image as AI Threats Grow, Guardian (Apr. 27, 2026), https://www.theguardian.com/music/2026/apr/27/taylor-swift-trademarks-voice-image-ai; Jay Peters, Taylor Swift Is Trying to Trademark Her Voice to Fight AI Copycats, Verge (Apr. 2026), https://www.theverge.com/ai-artificial-intelligence/919827/taylor-swift-trademarks-ai-copycats.
2. See 17 U.S.C. § 102(a); Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 345 (1991).
3. See In re Gen. Elec. Broad. Co., 199 U.S.P.Q. (BNA) 560, 563 (T.T.A.B. 1978).
4. See KP Permanent Make-Up, Inc. v. Lasting Impression I, Inc., 543 U.S. 111, 117 (2004).