From Consumer Confusion to Identity Protection: What Taylor Swift’s Trademark Strategy Reveals About Trademark Law in the Age of AI Deepfakes

Introduction : Consider using social media and listening to someone who has the voice, the cadence, the style of Taylor Swift and endorsing a product that she never used or a song that she never wrote. Millions listen, share and believe it. Days later, one finds that the whole audio was created by AI, without any recording, no studio, no contract signed. So, the question is if no original Taylor Swift recording was copied, then what is being infringed?

In January 2024, sexually explicit deepfakes of Taylor Swift appeared on social media, including one that allegedly had over 47 million views before it was removed from X (formerly Twitter). The incident ignited a national and international discussion on the extent of existing law, resulted in a U.S. Senate hearing and caused concern from the White House.

However, the deepfake issue is far from being limited to explicit videos without consent. The same technology can be used with no effort to create a realistic AI voice that promotes a financial product, a fake video of a brand’s launch, or a fake track to sound like a brand’s new release. In each of these scenarios, the same legal question arises: which branch of intellectual property law does it apply to?

The answer, more and more, is trademark law, but not in a way in which it was intended to be used. As AI-generated deepfakes become more prevalent, copyright and personality-rights laws are proving inadequate to protect the identity of celebrities and their counsel, leading to a new form of protection: trademark. With the proliferation of AI-generated deepfakes, celebrities and their counsel are turning to trademark protection as an untapped and unexpected way to safeguard their identities.

Struggles of Traditional IP Rights

A. An identity cannot be protected by copyright

A song, a sound recording, a film performance: all are protected by copyright law. It does not safeguard the identity of a person. A flawlessly cloned voice is one that is created entirely by an artificial intelligence, without taking any frame of the original content. There is no copying in the sense of copyright. No protected work is being reproduced, only an imitation is being made.

When an AI system produces a new recording that sounds something like Taylor Swift, but isn’t her actual work, the element of a protected work being reproduced is missing from the copyright equation. Some U.S. lawmakers have introduced legislation such as the NO FAKES Act and the DEFIANCE Act that would establish new rights for digital replications. At this time, these bills are in the proposal stage and have not been passed into law.

B. Personality Rights in India are Judicially Recognised but Legislatively Incomplete

Personality rights in India are not defined by any separate statute but rather are part of the general law of Intellectual Property and include the right of a public figure to control the commercial use of his name, image, voice and likeness. These have been evolved over time by judicial interpretation of Article 21 of the Constitution (right to life and personal dignity).

The judgment in Amitabh Bachchan v. Rajat Nagi & Ors. by the Delhi High Court issued an omnibus injunction that bars the world at large from using the name, image, voice or any attribute that is uniquely associated with the actor for any commercial or personal gain. The court ruled that the defendants were exploiting Bachchan’s popularity to sell products and services without his consent, thereby causing “irreparable harm” to his “identity and reputation”.

A year later, in Anil Kapoor v. Simply Life India & Ors., the Delhi High Court issued the first court order in India that specifically prohibited 16 entities from using Kapoor’s name, likeness, image and voice for monetary or other benefit, and from using technological tools like AI, ML, face morphing, and GIFs. Justice Prathiba Singh noted that celebrity also enjoy right to privacy and does not want his or her image, voice being used in a “dark manner” as it is being done on porn websites.

These decisions are important but they also shed light on the boundaries of the personality-rights approach in India. The relief was based on a mix of constitutional rights, common law passing off and copyright, not a particular statute. In the absence of codified personality-rights law, every case must be considered on its own merits and results are unpredictable from jurisdiction to jurisdiction.

This results in a protection gap. If the AI creates a voice that resembles a celebrity but hasn’t copied their voice, there is no copyright. Certainty is limited when personality rights need to be developed by the courts rather than being codified by the legislature and Trademark law is starting to fill this space.

Entry of Trademark Law in the Conversation of AI Deepfakes and Identity Protection

A. The Traditional Function of Trademark Law

The purpose of trademark law is to serve consumers and markets. A trademark is a word, name, symbol or device that designates the source of goods or services, distinguishes them from other goods and services, and provides protection to the goodwill invested in a brand by the producer. The basic intent of the Trade Marks Act, 1999 (India) and similar laws in other jurisdictions is to avoid causing any confusion in the marketplace regarding the commercial origin of goods and services.

The use of a registered trademark, by a person other than the registered proprietor or a permitted user, in the course of trade of a mark that is identical with or deceptively similar to the registered trademark, and that is likely to cause confusion or that takes unfair advantage of or is detrimental to the distinctive character or repute of the mark is an infringement of the trademark. Section 27(2) also retains the common law passing off rights with respect to unregistered marks.

B. Trademark Law’s Emerging Function: Protecting Identity, Not Just Products

Trademark rights are now being asserted beyond the context of product branding, in the case of celebrity identification, to protect the commercial value of the celebrity’s name, voice, and persona as a source-identifier.

The mechanism is false endorsement, under the law of passing off. When the ad appears to be a celebrity’s voice or likeness endorsing, approving of, or being associated with a product, it leaves the consumer with a false impression that the celebrity has endorsed, approved of, or is associated with that product. The misrepresentation is the very sort of market deception that was the primary intent of the trademark law.

The principle was established by the Indian courts as early as in the case of D.M. Entertainment Pvt. Ltd. v. Baby Gift House and Ors. Delhi High Court had ruled in favour of the singer Daler Mehndi in the case of unauthorised use of his personality for commercially sold dolls. Unauthorised use of Mehndi’s name and likeness was held to be false endorsement and passing off by the court, which found that the defendant had misappropriated a persona that had “grown to the status of a quasi property right” with commercial value. Importantly, the court considered Section 29 of the Trade Marks Act, 1999 and found that it is an infringement to use a celebrity’s name and mark in a deceptively similar way without permission in the course of trade.

Trademark law is no longer just asking if consumers are confused about products. Rather it is being asked whether consumers are confused about people and whether a person has endorsed or approved what he or she is seeing and hearing. That’s a big shift in doctrine.

Taylor Swift as a Case Study in Strategic IP Protection

A. Commercial Identity of Celebrity

Taylor Swift isn’t just an artist. Her name, her voice, her initials, her song names, her fan community slang, and even her cats’ names have been protected as legal assets. She is brand architecture, meticulously built by her for over 20 years of careful and strategic trademark protection.

B. A Trademark Portfoli Designed for the Digital Era

Swift’s trademark strategy is a case study in the protection of celebrity identity in the age of AI. As part of her business, TAS Rights Management LLC, Swift has amassed one of the largest celebrity trademark portfolios in the United States. Her first trademark applications were made in March 2007 and the first trademark for “TAYLOR SWIFT” was registered in 2008. Since then, she has applied for more than 350 trademarks and now has more than 150 active registrations across a wide range of international trademark classes.

Her portfolio includes her name “TAYLOR SWIFT” registered in multiple classes of merchandise and entertainment goods, her initials “T.S.” in over 21 design variations corresponding to different album eras, album titles such as FEARLESS, 1989 and REPUTATION, fan community terms like “SWIFTIE” (U.S. Trademark Registration No. 6,646,524) and “SWIFTIES” (U.S. Trademark Registration No. 6,646,529), the phrase “TAYLOR’S VERSION” (U.S. Trademark Registration No. 7,844,575), as well as her signature as a stylised design mark. In 2023, “TAYLOR SWIFT TOURING” was registered for live event services, which, as you might imagine, happened to coincide with the Eras Tour.

It is important to understand what this portfolio will do: registration will give a rebuttable presumption of ownership, establish a priority date, and give the registered proprietor a statutory cause of action for infringement. These registrations serve as a form of proactive legal framework for a celebrity in the age of AI. They are not just defense weapons but also the basis for enforcement.

C. What do Deepfakes Change About This Equation?

AI redefines a celebrity’s identity as a replicable asset. If a generative AI model is trained on publicly available audios, it can create a voice that is, to most listeners, identical to the original. No recording is duplicated. No image is stolen. But the commercial damage is the same and often more severe than the classic misappropriation because the amount of dissemination is vastly greater and the obstacles to creation are down to almost nothing.

In this context, trademark law is especially appealing for three reasons. First, it’s about consumer perception, not about copying and AI-generated content is, in fact, a manipulation of perception. Second, a registered trademark is not tied to whether or not original creative content was reproduced, but whether a protected identifier was used in commerce without authorization in a manner that is confusing. Third, the passing off doctrine of trademark law which has three elements (the goodwill of the trademark owner, misrepresentation regarding endorsement or association, and damage) is well-suited to the harm of a deepfake. The goodwill of the celebrity exists, the misrepresentation is generated by the AI, and damage follows, as the celebrity’s reputation and commercial interests are harmed.

Can Trademark Law in India handle AI Deepfakes?

This is the question that needs to be taken seriously in India by practitioners and businesses.

A. The Existing Framework

The Trade Marks Act, 1999 offers a viable (although imperfect) base. To make celebrity names registrable as trademarks, the Act defines “mark” as including names in section 2(1)(m). A use action for infringement under Section 29 is available for use of identical or deceptively similar marks in trade. The protection of well-known trademarks under Section 29(4)(c) is reserved for trademarks that are well known in India, which are used in a manner that is unfair and/or detrimental to the distinctive character of the mark, even if the goods or services are different. Section 27(2) maintains the passing off doctrine as a common law remedy, even for unregistered marks.

The false endorsement doctrine is already part of the passing off doctrine in India, as judged by the D.M. Entertainment judgment. The Anil Kapoor order is a testament to the Court’s willingness to give a special protection to AI-generated deepfakes of celebrities’ personas.

B. Applying the Framework to the Deepfake Scenario

Let’s take an example: An AI-generated audio segment includes a recognizable Indian celebrity, such as a cricketer who has trademarks for their name and signature, delivering a seemingly endorsement for a financial product. The video is shared on social media, and has been viewed millions of times. No original recording was used. There is no contract of endorsement.

The lawyer for the celebrity would argue that: (i) the celebrity’s name is a registered trademark; (ii) the AI-generated clip uses the name and/or an indicia of the celebrity's identity in the course of trade; (iii) this is likely to lead to consumer confusion, and is exactly what Section 29 is meant to prevent; and (iv) passing off applies independently of a trademark, as the celebrity has established goodwill, the clip is a misrepresentation, and as a result reputational and commercial harm is suffered.

The practical uncertainty is in the application and not so much in the principle itself. The courts will have to determine the followings:

• what will constitute as “use in the course of trade” in relation to the social media content created with AI;

• whether the absence of the registered name itself (where only the voice or likeness is used) is enough; and

• how to identify and serve the defendants, where the creator of the deepfake may be anonymous or not present in India.

C. The Legislative Gap

The IT (Amendment) Act, 2023 had inserted Rule 3(1)(b)(vii) in the IT Rules, 2021 which prescribed that the social media intermediaries take adequate measures to ensure that the users do not impersonate another person. A positive but limited measure. It holds platforms accountable but not the wrongdoers who are responsible for the creation of deepfakes and other harms. In addition, the enforcement is still in its infancy. There’s no specific deepfake law, as compared to U.S. states, and the framework relies on the judicial interpretations of Courts.

Quiet Transformation from Consumer Confusion to Identity Protection

One thing that's important is to make a clear statement of the doctrinal shift that's happening. The main goal of the trademark law was to control markets; to protect consumers from being misled about the commercial origin of products; and to protect producers from losing their investment in brand goodwill. The market was its beneficiary.

AI deepfakes are essentially, in a sense, performing an act that trademark law was never meant to protect against the commercial use of someone's name without any copying.

The issue is whether this is a valid application of the trademark doctrine or a stretch. The case for valid extension seems to be strong; given a false endorsement via a deep fake is an economic and reputation harm that is the same as passing off. The process of consumer confusion (association) is the same. Only the technology is different and Trademark law is not necessarily more concerned with the technology than with the harm it causes.

But there is also a strong argument for overreach: Trademark protection of identity can be a danger to legitimate commentary, parody, and creative expression. As this area develops, courts will have to develop more sophisticated defences for fair use and public interest.

What is increasingly clear is that celebrity, celebrity management and celebrity counselors are not waiting for legislation. Their aim is to create trademark portfolios that are designed to not only defend merchandise and tour branding but also to set the legal infrastructure that can be used to fight back against AI-generated impersonation in courts, specially when the courts are more familiar with trademark law instead of new AI-specific regimes.

Conclusion

AI deepfakes are not just causing new types of damage, they are also sparking new legal changes. The strategy that Taylor Swift has adopted, conscious and long-term and encompassing all facets of her commercial brand, is a new application of trademark law to reduce harms that were never envisioned to be the responsibility of personality-rights law.

The Indian courts have already shown their willingness to curb deepfakes by means of personality rights and passing off, as seen in the Anil Kapoor and Amitabh Bachchan cases. However, the road to be taken is rough. The effectiveness and reliability of Indian trademark law in preventing the use of AI-generated misrepresentation of a celebrity's identity will be subject to judicial interpretation of “use in the course of trade”, the judicial interpretation of voice and likeness as a sign of a registered mark, and the possibility of Parliament enacting specific legislation.

What’s clear is that, in the age of AI deepfakes, the issue trademark lawyers and their clients will face is not just whether a brand belongs to someone. It's about who has an identity and what it takes to preserve it.

Author: Jyotsana Singh, in case of any queries please contact/write back to us via email to chhavi@khuranaandkhurana.com or at  Khurana & Khurana, Advocates and IP Attorney.

Citations and References

1. Trade Marks Act 1999.

2. Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules 2021.

3. Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2023.

4. Anil Kapoor v Simply Life India & Ors CS(COMM) 652/2023 (Delhi High Court, 20 September 2023).

5. Amitabh Bachchan v Rajat Nagi & Ors CS(COMM) 819/2022 (Delhi High Court, 25 November 2022).

6. D M Entertainment Pvt Ltd v Baby Gift House and Others CS(OS) 893/2002 (Delhi High Court, 2010).

7. R Rajagopal v State of Tamil Nadu (1994) 6 SCC 632.

8. Titan Industries Ltd v Ramkumar Jewellers 2012 SCC OnLine Del 2382.

9. United States Patent and Trademark Office, ‘SWIFTIE’, US Trademark Registration No 6646524 https://tsdr.uspto.gov/#caseNumber=87337351 accessed 29 June 2026.

10. United States Patent and Trademark Office, ‘SWIFTIES’, US Trademark Registration No 6646529 https://tsdr.uspto.gov accessed 29 June 2026.

11. United States Patent and Trademark Office, ‘TAYLOR’S VERSION’, US Trademark Registration No 7844575 https://tsdr.uspto.gov/#caseNumber=90523859 accessed 29 June 2026.

12. United States Patent and Trademark Office, ‘SWIFTMAS’, US Trademark Registration No 5697658 https://tsdr.uspto.gov/#caseNumber=86839024 accessed 29 June 2026.

13. United States Patent and Trademark Office, ‘Trademark Search’ https://tmsearch.uspto.gov accessed 29 June 2026.

14. Hannah Dailey, ‘Taylor Swift’s Trademark List: From “Swiftie” to “Taylor’s Version” Billboard (updated 2026) https://www.billboard.com/lists/taylor-swift-trademark-list-swiftie-taylors-version/ accessed 29 June 2026.

15. ‘They Belong with Taylor Swift’ (2020) Illinois Law Review Online https://illinoislawreview.org/online/they-belong-with-taylor-swift/ accessed 29 June 2026.

16. Vanderbilt Law School, ‘Taylor Swift & AI Deepfakes: Swifties Join the Fight Against Deepfake Pornography’ (February 2024) https://law.vanderbilt.edu/taylor-swift-ai-deepfakes-swifties-join-the-fight-against-deepfake-pornography/ accessed 29 June 2026.

17. Julia Sturges, ‘Taylor Swift, Deepfakes, and the First Amendment: Changing the Legal Landscape for Victims of Non-Consensual Artificial Pornography’ (2024) Georgetown Journal of Gender and the Law (Online Edition) https://www.law.georgetown.edu/gender-journal/online/volume-xxv-online/taylor-swift-deepfakes-and-the-first-amendment-changing-the-legal-landscape-for-victims-of-non-consensual-artificial-pornography/ accessed 29 June 2026.

18. ‘Why Taylor Swift’s AI Scandal Is Pushing Lawmakers to Address Pornographic Deepfakes’ National Law Review (April 2024) https://natlawreview.com/article/why-taylor-swift-ai-scandal-pushing-lawmakers-address-pornographic-deepfakes accessed 29 June 2026.

19. World Intellectual Property Organization, ‘Taylor Swift Trademark Strategy: A Model for Artist IP Protection’ (19 August 2025) https://www.wipo.int/en/web/wipo-magazine/articles/taylor-swift-trademark-strategy-a-model-for-artist-ip-protection-78728 accessed 29 June 2026.