WHO OWNS A VOICE?: TAYLOR SWIFT’S TRADEMARK STRATEGY AGAINST AI DEEPFAKES: Rethinking IP Protection for Celebrity Identity in the Age of AI

INTRODUCTION

Artificial intelligence has fundamentally altered the economics of identity. Generative tools are now capable of replicating an individual’s voice, speech cadence, facial likeness, and performance style with alarming precision, often using minimal publicly available data. While such technologies offer legitimate creative and commercial opportunities, they simultaneously create unprecedented legal vulnerabilities for public figures whose identities function as valuable commercial assets.

This tension became especially visible in April 2026, when Taylor Swift reportedly filed three trademark applications before the United States Patent and Trademark Office (“USPTO”), seeking registration over two sound identifiers: “Hey, it’s Taylor” and “Hey, it’s Taylor Swift”, alongside a signature image associated with her Eras Tour persona.[1] These filings were widely interpreted as a strategic legal response to rising concerns over AI-generated deepfakes, voice cloning, and unauthorized digital impersonation.

Swift’s filings signal more than a routine celebrity branding exercise. They illustrate an emerging shift in intellectual property strategy: as generative AI increasingly replicates not merely creative works but creators themselves, celebrities are expanding trademark law into areas traditionally governed by copyright and personality rights.

This article argues that conventional copyright law is structurally inadequate to protect celebrity voice and likeness from AI-driven imitation. While trademark law and publicity rights provide partial safeguards, Taylor Swift’s recent filings demonstrate both the utility and limitations of existing frameworks, ultimately underscoring the need for AI-specific identity protections.

AI DEEPFAKES AND THE COMMERCIAL THREAT TO CELEBRITY IDENTITY

AI-generated deepfakes have transformed impersonation from a niche novelty into a scalable commercial risk. Unlike traditional parody or imitation, generative AI can now produce highly convincing vocal and visual replicas capable of misleading consumers, damaging reputations, and displacing authentic creative output.

In 2023, an AI-generated song titled Heart on My Sleeve, featuring synthetic imitations of Drake and The Weeknd, went viral before being removed following intervention by Universal Music Group.[2] The controversy highlighted a novel legal issue: while the underlying recording was AI-generated, the commercial value derived largely from its imitation of recognizable artist identities.

Similarly, in early 2024, AI-generated explicit images falsely depicting Taylor Swift circulated widely across social media platforms, generating public backlash and reigniting policy debates concerning non-consensual digital replicas.[3] Though image-based, the incident illustrated a broader structural problem, public figures currently possess limited legal tools to prevent unauthorized synthetic reproduction of their identity.

For artists, identity is economically inseparable from output. A singer’s voice is not merely biological expression; it is a commercially monetizable asset tied to endorsements, consumer trust, licensing, and brand value.

Unauthorized AI replication can therefore generate multiple harms:

• reputational injury through false associations;
• consumer deception regarding endorsements or releases;
• market substitution by competing synthetic content.

These harms reveal a widening gap between technological capability and legal preparedness.

WHY COPYRIGHT LAW FAILS TO PROTECT VOICE AND PERSONA

Despite its centrality to creative industries, copyright law offers limited protection against AI voice cloning and identity replication.

Under the Copyright Act of 1976, copyright subsists in original works of authorship fixed in a tangible medium, including musical compositions and sound recordings[4].^4 However, copyright does not extend to an individual’s voice, vocal style, or mannerisms.

This limitation follows directly from the idea-expression dichotomy. This limitation is reinforced by 17 U.S.C. § 102(b), which clarifies that copyright protection does not extend to ideas, procedures, systems, methods, or similar non-expressive elements. Copyright protects the expression of ideas, not the ideas, methods, or styles themselves. Thus, while a specific Taylor Swift recording is protected, an AI model generating a new recording that merely sounds like her may avoid infringement absent copying of protected expression. The inadequacy of copyright in protecting voice identity is well illustrated by judicial precedent.

In Midler v. Ford Motor Co., singer Bette Midler successfully challenged the use of a sound-alike vocalist in a commercial after refusing to participate herself.[5] Notably, the claim did not succeed under copyright law, but through tort-based misappropriation principles.

Likewise, in Waits v. Frito-Lay, Inc., Tom Waits prevailed after a company deliberately imitated his distinctive voice in advertising.[6] Again, liability arose not from copyright infringement, but false endorsement and commercial appropriation. Further, the U.S. Copyright Office explicitly states that short phrases, names, slogans, and titles are not copyrightable.[7] Consequently, phrases such as “Hey, it’s Taylor” cannot ordinarily be protected through copyright, regardless of their association with a celebrity.

This doctrinal architecture reveals a fundamental mismatch: copyright protects creative works but not the commercially valuable identity traits through which such works are recognized.

III. TAYLOR SWIFT’S TRADEMARK STRATEGY: EXPANDING TRADEMARK INTO IDENTITY PROTECTION

Against this backdrop, Taylor Swift’s April 2026 trademark filings represent a strategically sophisticated use of trademark law.

Under the Lanham Act, a trademark includes any word, name, symbol, device, or combination thereof used to identify and distinguish goods or services and indicate their source.[8] By filing applications over “Hey, it’s Taylor” and “Hey, it’s Taylor Swift” as sound marks, Swift appears to be transforming recurring auditory identifiers into legally protectable commercial assets.
As these applications remain pending at the time of writing, their legal significance is currently strategic rather than dispositive.

This strategy is doctrinally significant for two reasons.

First, trademark law may offer stronger enforcement tools where AI-generated replicas create consumer confusion or false endorsement concerns. If consumers encounter synthetic content using protected sound identifiers in a manner suggesting association with Swift, trademark claims may become viable.

Second, the filings suggest a broader expansion of trademark beyond conventional branding into defensive identity protection.

Historically, celebrity trademarks have focused on merchandise, lyrics, album titles, or logos. Swift herself has previously filed trademarks connected to album branding and phrases linked to her musical eras.[9] However, the 2026 filings indicate a more targeted anti-deepfake rationale: protecting recognizable vocal and visual markers before AI misuse escalates.

This reflects an evolving legal reality in which celebrities increasingly treat identity itself as trademarkable infrastructure.

Swift’s inclusion of a signature image mark alongside sound marks reinforces this strategy. In effect, the filings construct a layered IP shield over both auditory and visual components of her persona.

This is not merely brand management, it is anticipatory legal architecture.

THE LIMITS OF TRADEMARK AND THE NEED FOR AI-SPECIFIC REFORM

Despite its strategic utility, trademark law is not a complete solution.

Trademark protection depends on proving source-identifying use and likelihood of confusion in commerce.[10] This leaves substantial gaps where AI-generated replicas are distributed non-commercially, anonymously, or outside conventional trademark contexts.

Moreover, trademark strategies disproportionately benefit high-profile celebrities with substantial financial and legal resources. Emerging artists, creators, and ordinary individuals remain significantly less capable of proactively trademarking identity markers.

The right of publicity offers a more conceptually direct framework by protecting against unauthorized commercial exploitation of identity.

In Carson v. Here’s Johnny Portable Toilets, Inc., television host Johnny Carson successfully restrained commercial appropriation of his persona even where his literal name was not used.[11] Similarly, California law explicitly protects voice as an aspect of identity.[12]

Indian courts have recognized limited publicity and personality rights through judge-made jurisprudence, though India lacks a comprehensive statutory framework addressing AI-generated identity misuse. In D.M. Entertainment Pvt. Ltd. v. Baby Gift House, the Delhi High Court acknowledged publicity rights concerning singer Daler Mehndi and restrained unauthorized commercial exploitation of his persona.[13]

Yet publicity rights remain fragmented, jurisdiction-specific, and largely underdeveloped for AI contexts. This fragmentation underscores the need for targeted legal reform.

Legislatures should consider:

• statutory recognition of digital voice and likeness rights;
• consent requirements for commercial AI replication;
• platform obligations concerning synthetic identity disclosures;
• civil remedies against unauthorized deepfake deployment.

Recent U.S. proposals such as the No AI FRAUD Act indicate growing legislative recognition of these concerns.[14] However, existing reforms remain embryonic.

Absent AI-specific intervention, intellectual property law risks protecting songs while leaving singers vulnerable.

CONCLUSION
Taylor Swift’s April 2026 trademark filings reflect a significant evolution in celebrity intellectual property strategy. By seeking protection over sound identifiers and image-based branding elements amid rising AI concerns, Swift is not merely expanding her commercial portfolio, she is responding to a legal vacuum.

Copyright law remains structurally incapable of protecting voice, style, and persona from synthetic imitation. Trademark law and publicity rights offer partial workarounds, but their fragmented scope and doctrinal limits prevent comprehensive protection.

As artificial intelligence increasingly reproduces not only works but workers, artists, performers, and public figures, the future of intellectual property law may increasingly require stronger recognition of commercially valuable digital identity interests. In this emerging landscape, Taylor Swift’s filings may prove less a celebrity anomaly than an early blueprint for the next generation of intellectual property protection.

BIBLIOGRAPHY

1. Copyright Act of 1976, 17 U.S.C. §§ 101–1401 (2024).
2. Lanham Act, 15 U.S.C. §§ 1051–1141n (2024).
3. Cal. Civ. Code § 3344 (West 2024).
4. Carson v. Here’s Johnny Portable Toilets, Inc., 698 F.2d 831 (6th Cir. 1983).
5. D.M. Ent. Pvt. Ltd. v. Baby Gift House, 2010 SCC OnLine Del 479.
6. Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988).
7. Reuters, Taylor Swift Files Trademark for Her Voice and Likeness to Ward Off AI Deepfakes(Apr. 27, 2026).
8. Rio Yamat, Fake Explicit Taylor Swift Images Renew Concerns Over AI Harms, Associated Press (Jan. 26, 2024).
9. U.S. Copyright Office, Circular 33, Works Not Protected by Copyright(2021).
10. Waits v. Frito-Lay, Inc., 978 F.2d 1093 (9th Cir. 1992).
11. Ben Sisario, An AI-Generated Song Went Viral. Music Labels Were Not Amused, N.Y. Times (Apr. 19, 2023).
12. Maria Sherman, Taylor Swift Files New Trademark Applications. One Expert Says It Is to Curb AI Threats, PBS NewsHour (Apr. 28, 2026).

[1] Reuters, Taylor Swift Files Trademark for Her Voice and Likeness to Ward Off AI Deepfakes (Apr. 27, 2026), https://www.reuters.com/legal/litigation/taylor-swift-files-trademark-her-voice-likeness-ward-off-ai-deepfakes-2026-04-27/

[2] Ben Sisario, An AI-Generated Song Went Viral. Music Labels Were Not Amused, The New York Times (Apr. 19, 2023) https://www.nytimes.com/2023/04/19/arts/music/ai-drake-the-weeknd-fake.html

[3] Deepfake Explicit Images of Taylor Swift Spread on Social Media, The Hindu (Jan. 27, 2024), https://www.thehindu.com/entertainment/movies/deepfake-explicit-images-of-taylor-swift-spread-on-social-media/article67782252.ece

[4] Copyright Act of 1976, 17 U.S.C. §§ 101–1401

[5] Midler v. Ford Motor Co., 849 F.2d 460, 467 (9th Cir. 1988).

[6] Waits v. Frito-Lay, Inc., 978 F.2d 1093, 1103 (9th Cir. 1992).

[7] U.S. Copyright Office, Circular 33: Works Not Protected by Copyright (2021), https://www.copyright.gov/circs/circ33.pdf

[8] Lanham Act, 15 U.S.C. §§ 1051-1141n (2024)

[9] Maria Sherman, Taylor Swift Files New Trademark Applications. One Expert Says It Is to Curb AI Threats, PBS NewsHour (Apr. 28, 2026), https://www.pbs.org/newshour/arts/taylor-swift-files-new-trademark-applications-one-expert-says-it-is-to-curb-ai-threats

[10] 15 U.S.C. § 1114 (2024).

[11] Carson v. Here’s Johnny Portable Toilets, Inc., 698 F.2d 831 , 835 (6th Cir. 1983)

[12] Cal. Civ. Code § 3344 (West, 2024)

[13] D.M. Ent. Pvt. Ltd. v. Baby Gift House, 2010 SCC OnLine Del 479.

[14] No AI FRAUD Act, H.R. 6943, 118th Cong. (2024), https://www.congress.gov/bill/118th-congress/house-bill/6943