AI can imitate a singer's recognisable voice without copying the original recording, exposing a gap between copyright and protection of vocal identity.
In Arijit Singh v. Codible Ventures LLP (2024), the Bombay High Court considered the unauthorised use of the singer's voice and manner of singing. The Delhi High Court's 2025 order in Aishwarya Rai Bachchan v. Aishwaryaworld.com & Ors. similarly illustrates judicial willingness to grant interim protection against unauthorised exploitation of personality attributes, including through artificial intelligence. The question is which right, if any, is infringed when a synthetic song imitates a performer without reproducing an existing recording.
Copyright's Structural Limitations
In the Copyright Act, 1957, there is specific protection accorded to different subject matters via rights in literary and musical works, sound recordings and performances. Sections 38, 38A and 38B are concerned with performer's rights, exclusive rights in performances and performers' moral rights respectively. But in terms of legal protection, the Copyright Act, 1957, does not provide for a proprietary right to each and every individual characteristic of one's voice.
If the artificial intelligence generates a new sound recording which imitates distinctive voice characteristics but does not reproduce the recording or performance of the singer, then the question is whether or not there could be copyright infringement in such a scenario. While copying an existing sound recording would amount to copyright infringement based on statutory provisions, the other scenario requires analysis of both the output and training material before making a determination of copyright infringement.
This poses difficulties in relation to Sections 38 and 38A in a situation where no recording or performance protected by copyright law is reproduced. Section 38B is concerned with attribution and prejudicial distortions or modifications to performance and should not be resorted to as an easy solution to any independently generated imitation.
The Training Stage: A Separate Copyright Question
Scrutiny must look beyond the final song at the training material and processes for use of voice-cloning systems. Such systems may use copyrighted recordings to learn speech phonation, rhythm, pronunciation and other vocal features. Depending on the conduct involved, reproducing, storing or processing protected recordings may engage the exclusive rights under Section 14 of the Copyright Act.
Section 52 contains various specified exceptions, but Indian law does not have a comprehensive, AI-specific text- and data-mining exception which explicitly authorizes all commercial AI training. Whether a particular practice falls under an existing exception depends on the nature of the use, acts carried out, as well as applicable statutory conditions. The mere fact that the use is for a commercial purpose is not sufficient evidence of non-infringement, but calling such conduct training does not make it exempt from being considered.
AI-generated music thus poses two separate copyright questions: whether acquiring and using the training material infringed or not, or fell within an applicable exception, and whether the resulting recording reproduces protected expression, or otherwise violates an existing right. A non-infringing output does not establish that the training was lawful. Conversely, use of copyrighted material during training does not guarantee that every output infringes copyright.
Personality Rights as an Alternative Avenue
Personality rights shield attributes, such as name, likeness and voice, that are uniquely identified with a person from unlawful use for commercial ends. Their scope is defined by the applicable doctrine and factual matrix and they do not provide absolute control over every form of duplication of an individual's style. In Titan Industries Ltd. v. Ramkumar Jewellers (2012), the Delhi High Court took note of the commercial value of the celebrity's identity when it came to the unauthorized use of his name in advertisements and the proceedings concerning Amitabh Bachchan vs Rajat Nagi & Ors on 25 November, 2022 and Anil Kapoor vs Simply Life India & Ors on 20 September, 2023, among others, have demonstrated the courts' receptiveness to granting remedies as relief against a variety of unauthorized uses of attributes, including those aided by technology. This does not mean, however, that every instance of resemblance would be actionable on account of the presence of identity-based claims.
The constitutional underpinnings of such an argument have also been considered in K.S. Puttaswamy (Retd.) v. Union of India (2017) and the Supreme Court's recognition of privacy as a fundamental right under Article 21 of the Constitution provides an additional dimension to personality rights beyond a mere commercial one.
The proceedings in Arijit Singh before the Bombay High Court pertain to an alleged use of the plaintiff's personality attributes – his voice and singing style – in an unauthorized manner. Such an interlocutory order is significant as it may offer remedies concerning personality-based rights, including those that go beyond the mere duplication of a song, for the technology that is available today allows the synthesis of performances that are visually and acoustically indistinguishable from the original. It is not necessarily a definitive statement on the existence of an exclusive property in the attribute of voice since the orders passed are preliminary in nature. Beyond the mere confusion that the use of another's style may cause the public, such acts can also impinge on the ability to control the licensing of one's works, affect an individual's professional relationships and tarnish their reputation and it is on the question of whether the effect of such unauthorized use should be considered a violation of the law that the courts in India appear to be taking a more receptive view.
It is a matter of determining when such use should be considered a violation of the law that the courts in India appear to be taking a more receptive view.
Is Disclosure Sufficient?
A difficult case would be a synthetically generated song which is clearly labelled as such and does not falsely suggest that the original singer performed or endorsed it. Such disclosure would undercut a passing-off claim by negating the requirement of misrepresentation, but would not necessarily defeat a distinct claim concerning unauthorised commercial exploitation of personality attributes, if the relevant legal requirements are satisfied.
A platform could brand a fake one, and then use it commercially without permission. However, the problem may not be deception but appropriation – disclosure and authorisation are not the same.
A deceptive replica, labelled commercial and non-commercial parody raises different issues. The mere possibility of identification should not be sufficient for liability; courts should weigh factors such as purpose, context, appropriation, harm and expressive protection.
Comparative Perspectives and the Need for Reform
Comparative models should be adopted in India. The ELVIS Act of 2024, passed in Tennessee, extended the list of prohibited uses of an individual's voice, including some AI avatars. For the same purpose, a federal law named NO FAKE Act of 2025 (S.1367) was introduced to the Senate on 9 April 2025, and it was referred to the Senate Judiciary Committee for further consideration on 11 April 2025. If passed, it would create a federal framework for regulating some unauthorized digital use of one's voice and visual image, which is significantly different from what was established in Tennessee.
The EU AI Act provides a parallel approach by obliging disclosure of certain AI generated or manipulated content, subject to relevant exemptions and exceptions. This is not the same as granting substantive rights of consent, remuneration or control over commercial exploitation. Disclosure highlights synthetic content but does not clarify whether the use of the identity is lawful.
The Indian position needs to consider a restrictive definition of a digital duplicate right. This right should seek to protect against unauthorised creation and exploitation of very realistic duplicates of a person's distinct voice. Consideration must be given to ensure that the protected duplicate(s) are substantially similar. It must be ensured that the right does not overlap with copyright, performers rights, passing off and current personality rights doctrine. The legislation should clearly define when consent is required and what remedies are available for actionable duplication.
Protections are required that ensure parody, satire, critique, news reporting, research and other legitimate uses are allowed through appropriate exemptions and remedies. Appropriate licensing and provenance records can be kept by commercial AI developers and platforms so that there will be genuine transparency into the creation of synthetic performances. This will help prevent misuse while also encouraging creative mimicry.
Therefore,
“Who owns a voice?” is a flawed legal question, as a voice, besides being the recorded sounds, can represent one's personality, professional reputation and even monetary value. Acknowledging these interests do not require having absolute authority over distinctive traits of another person's voice.
The cases of Titan Industries, the dispute between Amitabh Bachchan and Anil Kapoor, Arijit Singh and Aishwarya Rai Bachchan demonstrate how the existing legal doctrines can address the unauthorized use of certain aspects of someone's personality. Judicial interim measures are not sufficient in each case as the AI technology can generate performances which were not presented by the original actor.
As such, the Indian laws need to make the difference between an unlawful appropriation of an easily identifiable voice and legitimate artistic performance. A well-drafted digital replica doctrine which would secure expressive freedom, and strike the balance between technological progress and the actors' interests is needed. As the AI voice synthesis technology enables creating performances which were not presented by the real actors, the ability to imitate should not be confused with the right to do so. Such principles should be taken into consideration by courts, legislators, artists, developers and various digital platforms as the online music industry undergoes rapid changes at present.
Authors are 3rd year Law students at Indian Institute of Management Rohtak. Views are personal.