Caught In The Middle: When Your Vocal Identity Fall Through The Gaps In IP Law

Written by: Jeny Sara George, B.A.LL.B (Hons.) | Christ University Pune Lavasa 

JUSTICE

In early 2023, a song titled “Heart on my sleeve” went viral not because of the lyrics in the song or the composition of the song but because it sounded uncannily like the musicians Drake and The Weekend but neither of the artists recorded it. It was later understood that their voices were generated via artificial intelligence, which was trained on the existing voice data for it to produce something like this composition that is very eerily authentic[1]. Around this time, the same emerged in India about the replacement of the voices of musicians by AI. This instance or phenomenon exposed a legal paradox: modern IP law is readily available to protect a 30-second video clip under copyright or a specific brand logo or shade under trademark yet it remains a struggle to protect something far more personal, far more distinctive and the most primal signature of a person i.e the human voice.[2]

          These scenarios can be called a “voice land grab” where AI systems are trained on a massive data of recorded speeches and music and create a pattern extracting the same information to define one’s vocal identity. But the legality of the same with the current laws, which were designed for external creations like books, paintings, symbols, etc., and not for something fluid or more into biology like a person’s voice.[3] My blog argues that the current law creates a dangerous vacuum whereby falling between copyright’s requirement for originality and trademark’s requirement of source identification or commercial benefit, the human voice has become more of an unprotected side in this era.

Is A Personal Commodity, here ‘Voice’; A “Work[4]” Or A “Tool”?

Copyright law protects “original works of authorship that is fixed in a tangible medium”, also known as the Doctrine of Fixation. Under the Indian Copyright Act, 1957, at first glance, it seems that it is bad enough to include voice, but the law draws a very thin yet crucial distinction, protecting the expression, not the instrument that produces it. A voice is legally treated as an instrument or a biological by product than a creative expression of an individual. So, under Copyright law if the voice is treated as an instrument, then the expression created, i.e a song, or lyrics or a sound recording can be copyrighted. But the voice itself is not, the problem lies in the requirement of fixation, a voice is to be captured in a recording to be protected but even then, what is protected is the recording and not the quality of the voice. The uniqueness of the voice, the timbre, the pitch are all recognizable elements of a person’s voice that is outside the scope of copyright.  The Ninth Circuit said in the case of Midler v. Ford Motors [5]that this limitation becomes very evident where a singer was said to imitate Bette Midler’s voice in a commercial without the prior permission of the actual singer. The court in this case did acknowledge that voice is not copyrightable, deliberately imitating it could be said to be misappropriation of identity, hence it can be said that though this case did not expand this very limit it surely exposed this hidden limitation.

          In India, the Copyright law does provide performer’s rights under Section 38[6], 38A[7] and 38B[8]. These very rights protect performers’ personality and their specific performance from unauthorized use, but the very same law does not extend to the situation where an AI generates a new performance using the performer’s cloned voice. Even though no existing recording is copied, it can be said that there is no infringement at all. So, this means that while the output of the artist is protected by copyright law, but the essence of the artist is up for sale to the others[9].

Can A Human Be A Brand?

One of the most asked question is that if the copyright law fails to fully protect ones voice, can trademark law step in and be a saviour? Can it be protected under Non- Traditional Marks or Non- Conventional Marks?

          Trademarks law majorly protects the “source identifying” i.e signs that distinguishes goods or services in the marketplace. Like for example, we accept the fact the Nokia tune or the MGM Lion’s roar are trademarks as they tell the consumer exactly who is providing the service and hence theoretically it could be a “Sound Mark”. The Sound Mark is identified via majorly 2 conditions or tests i.e

1.    The Commerce Test – the voice or the use of voice is to be used to sell goods or services specifically.

2.    The Confusion Test – to win an infringement suit, there must be acquired distinctiveness and hence the likelihood of confusion is to be proved.

Now this is where the major problem arises, Most voices even the ones that are highly distinctive are not used as a trademark as a singer’s voice in a song to an actor's voice in a film does not function as a brand identifier legally as they aren’t necessarily “selling” a product nor is the public “confused” of the brand source. Even in cases where a voice becomes associated with a brand like narrator in the advertisements, the trademark protection again is extremely limited and by the time the voice becomes distinctive enough for trademark protection, the unauthorised AI training is already done. If an AI further generates a similar voice for a different purpose without misleading consumers about the origin then trademark law will not or may not provide a remedy as its not a commercial signaling but a vocal identity. This overlap is so narrow that the protection is conditional.

In India, a voice is considered as a performance under copyright or a mark under trademark but at the same time, a voice is not a performance or a mark, it is a part of identity, and such a scenario can lead to a failure. Indian courts have been constantly evolving the concept of Personality Rights and there are many judgements have been pro-celebrity. In the case of Amitabh Bachchan v. Rajat Negi[10], the Delhi high court issued an "omnibus" injunction where it protected Amitabh Bachchan's name, image, and persona from being used without his consent. The court also recognized that identity is also a commodity and should be protected. Similarly, in Anil Kapoor v. Simply Life India[11], the court protected that the Anil Kapoor's voice, image, his mannerism, and signature expression such as "Jhakaas". This case got it right in relation to the misuse of technology including AI driven impersonation which is a threat to personality rights.

Recently, In Arjit Singh v. Codbile Ventures LLP[12], the court granted interim relief restraining AI platform from the closing of singer’s voice, name or persona without consent, they also expressed their shock as to how the celebrity image is vulnerable to these generative tools. The Bombay High court also expressed their concern and granted similar relief to the singers Asha Bhosle, restraining the AI platforms from the use of voice cloning and that the generative tools facilitate fraudulent use of celebrities’ voice[13].

          While these can be considered as a win, these are defensive too. These verdicts are based on right of publicity which is not a statutory law in India but a judicial creation of right to privacy and hence can be termed as a "Rich Man's Remedy" as celebrities with means to fight get to protect their voices and an ordinary voice over artist and musician is not even legally visible[14].

The Crisis Due To AI

The most alarming issue for law is the training loophole for AI, where the AI models are trained for hours of vocal data in order to replicate speech patterns, mimic emotional tone, generate entirely new content in a person’s voice. Developers often justify this under “fair dealing” exceptions in Section 52 of the Indian Copyright Act[15] (narrower than the U.S. fair use doctrine), arguing that training is transformative. However, if the output i.e “The Synthetic Voice” can effectively replace a human artist in the market is it still fair in nature? Hence this raises concern of consent and economic exploitation and a broader societal risk of political misinformation, fraudulent communications, and Deepfake scams. Even though if the AI systems are trained on vast datasets and the training process is justified as “fair use” the output often replicate the style and identity of voices without copying the specific works.

Recommendations

Even though if the question “who owns a voice?” remains an unresolved question due to the existing law, trying to fit the voice in the existing boxes of copyright and trademark is not fully possible, they can be said that they are the 20^th-century tools to solve the 21 century biological crisis. Hence, I recommend a Sui Generis[16] Right for Vocal Identity where:

a.     Biometric Ownership- Unique tonal frequency of a human voice to be treated as personal property and should have personal property rights that exist beyond commercial use

b.    Making it illegal to train AI on a person’s voice without licensing agreement

c.     Protecting the voice as an identity and not as a work or a brand

Further steps, such as amending the Digital Personal Data Protection Act, 2024[17] and the Copyright Act, 1976, where voice data is sensitive personal data. India can also lead the world in protecting the voice at international level in forums like WIPO[18].

With technology making a void between human and machine voices, such protection becomes important. Allowing this to be in the way as it is a statement to the world that a logo is more valuable than the human soul in vocal form and in the 21st century, the law is the loudest you hear. As technology continues to blur in the boundaries between human and machine expression, the need for such protection becomes urgent. Letting this be in the way as it is telling the world that a logo is worth more than the human soul expressed through speech and in this era the loudest in the room is the law.

[1] Laura Snapes, AI Song Featuring Fake Drake and Weeknd Vocals Pulled from Streaming Services, The Guardian (Apr. 18, 2023), https://www.theguardian.com/music/2023/apr/18/ai-song-featuring-fake-drake-and-weeknd-vocals-pulled-from-streaming-services 

[2] Samantha Murphy Kelly, The Viral New ‘Drake’ and ‘Weeknd’ Song Is Not What It Seems, CNN (Apr. 18, 2023), https://edition.cnn.com/2023/04/18/tech/universal-music-group-artificial-intelligence 

[3] Debanjan Ranu, Hands Off My Voice: AI Voice Cloning and the Quiet Strength of Indian Law, IPVarna Research, https://ipvarna-research.in/ai-voice-cloning-personality-rights-indian-law/.

[4] Copyright Act, 1957, § 2(y) (India)

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

[6] Copyright Act, 1957, § 38 (India)

[7] Copyright Act, 1957, § 38A (India)

[8] Copyright Act, 1957, § 38B (India)

[9] Dipak G. Parmar, AI Voice Cloning: How a Bollywood Veteran Set a Legal Precedent, WIPO Mag. (Apr. 17, 2025), https://www.wipo.int/en/web/wipo-magazine/articles/ai-voice-cloning-how-a-bollywood-veteran-set-a-legal-precedent-73631 

[10] Amitabh Bachchan v. Rajat Negi & Ors., CS(COMM) 819/2022 (Delhi High Court 2022)

[11] Anil Kapoor v. Simply Life India & Ors., CS(COMM) 652/2023

[12] Arijit Singh v. Codmobile Ventures LLP, CS(COMM)

[13] Asha Bhosle v. Mayc Inc & Ors. (2025) SCC OnLine Bom 3485

[14] Titan Indus. Ltd. v. Ramkumar Jewellers, 2012 SCC OnLine Del 2382

[15] Copyright Act, 1957, § 52 (India)

[16] Ensuring Likeness, Voice, and Image Security (ELVIS) Act of 2024, Tenn. Code Ann. § 47-25-1101 et seq. (2024).

[17] Digital Personal Data Protection Act (DPDP), 2023 (India)

[18] Centre for Legal Innovation & Digital Society (CLINDS), Faculty of Law, Chinese Univ. of Hong Kong, CLINDS’s 26th Law & Digital Society Seminar: Vocal Identity Under Siege by AI Voice Cloning Technologies (Mar. 19, 2026), https://www.law.cuhk.edu.hk/app/events/clindss-26th-law-digital-society-seminar-vocal-identity-under-siege-by-ai-voice-cloning-technologies/.