AI Music and Copyright in India, Explained

AI music and copyright in India for indie artists: who owns AI-made songs, the DPIIT training plan, the ANI case, voice cloning and AI labelling rules.

If you make music in India and have opened an AI tool this year, you have probably wondered two things. Can I own a song I made with this? And what happens when someone uses a machine to copy me? The honest answer to both is that Indian law has started to move, but nothing about AI music and copyright in India is settled yet.

This guide sets out what is on the record as of October 2026, from the Copyright Act to the courts to what Spotify and YouTube now ask of artists. It is reporting on public documents, and it is not legal advice.

AI music and copyright in India at a glance

Authorship: the Copyright Act, 1957 already says the author of a computer-generated musical work is "the person who causes the work to be created"
Training: a DPIIT committee proposed in December 2025 that AI developers get a mandatory licence to train on lawfully accessed works and pay royalties. It is a proposal, not law
Courts: on 24 July 2026 the Delhi High Court refused ANI an interim injunction against OpenAI, in a case the major music labels backed. An appeal is pending
Voice: the Bombay High Court protected Arijit Singh's voice against AI cloning tools in July 2024, on personality rights
Labels: amended IT Rules in force since 20 February 2026 require platforms to label synthetic content, with a spoken disclosure at the start of synthetic audio

Who owns a song you made with AI?

Section 2(d)(vi) of the Copyright Act, 1957, as published on the Copyright Office's website, says that for any literary, dramatic, musical or artistic work "which is computer-generated, the person who causes the work to be created" is the author.

That clause predates generative AI by decades, and no Indian court has yet applied it to a song produced from a text prompt. The government committee studying the question was due to cover ownership of AI output in a second paper by February 2026. Hindustan Times reported on 5 October 2026 that it still had not appeared, and that three members said the panel had not met since June.

My view is simple. The more of the song that is plainly yours, the safer you are. A melody you wrote, lyrics you wrote and a vocal you sang are human work under any reading of the Act. A track generated end to end from one prompt is the open case. Keep your session files, voice memos and lyric drafts with their dates. If your authorship is ever questioned, that paper trail is the best evidence you will have.

Can AI companies train on your music?

On 9 December 2025 the Department for Promotion of Industry and Internal Trade published Part 1 of a working paper on generative AI and copyright, written by an eight-member committee it set up on 28 April 2025. According to the Press Information Bureau release, the committee rejected a zero-price licence, under which AI firms would train for free, and proposed what it calls a hybrid model. AI developers would get a blanket licence to train on all lawfully accessed content without negotiating with each owner. Royalties would become payable once the AI tool is commercialised, at rates set by a government-appointed committee and open to judicial review.

The paper itself is clear that this is a mandatory licence. Under the majority view, rights holders "will not have the option to withhold their works" from AI training. Money would be collected by a new central body, the Copyright Royalties Collective for AI Training, or CRCAT, made up of copyright societies, one per class of work. Non-members could be paid if they register their works, and the paper lists ISRC codes for recordings, ISWC codes for compositions and IPI numbers for music creators as identifiers that could be asked for.

Nasscom, which sat on the committee, dissented in a submission dated 17 August 2025, the paper records, preferring a text and data mining exception with a machine-readable opt-out for publicly available content.

Where does that leave you? DPIIT took public comments and extended the deadline to 6 February 2026, according to Hindustan Times, but no revised version has come out. I think the hybrid model would be better for independent musicians than an opt-out, because most of us will never know which crawler scraped what. It would only work, though, if the people being paid can be identified. An artist without ISRCs on her releases or an IPRS account for her songs would be hard to find in any royalty pool. Our guide to IPRS, PPL and Novex explains who collects for which layer of a song.

The ANI case, and why the labels joined it

India's first major court test of AI training is not about music, but the music industry made sure it was in the room. ANI sued OpenAI in the Delhi High Court in November 2024. The Federation of Indian Publishers, the Digital News Publishers Association and the Indian Music Industry body intervened in support of ANI, and Mint reported that T-Series, Saregama India and Sony Music India were the music companies behind that plea.

On 24 July 2026 Justice Amit Bansal refused ANI an interim injunction. According to the WIPO case summary of the order, the court held on a prima facie basis that OpenAI's storage of ANI's works for training fell within the fair dealing exception in Section 52(1)(a)(i) of the Copyright Act, which covers private or personal use, including research. It also found that the public interest weighed against an injunction, and said its observations were limited to the interim stage. The suit continues.

ANI appealed. On 15 September 2026 a Division Bench of Justice Avneesh Jhingan and Justice Manmeet Pritam Singh Arora issued notice to OpenAI. LiveLaw reported the next hearing as 8 December, while Business Standard gave 5 December.

I would not read the interim order as a green light for training on songs. It is a first look at news text, not a final ruling, and a lawyer quoted anonymously by Mint expected music companies to file their own suits, believing their claims are stronger.

When someone clones your voice

A fake version of you singing a song you never recorded is a more personal problem. The clearest Indian ruling on this is the Bombay High Court's order of 26 July 2024 in Arijit Singh v. Codible Ventures LLP. Justice R.I. Chagla granted ex parte interim relief against several defendants, including AI platforms. The order records the singer's case that a data set of 456 of his songs had been uploaded to one platform so users could convert any voice into his.

The judge's words are worth quoting exactly: "Making AI tools available that enable the conversion of any voice into that of a celebrity without his/her permission constitutes a violation of the celebrity's personality rights."

The catch for independent artists is in the same order. It describes establishing the plaintiff's celebrity status as "only the primary ingredient" of a personality rights claim, alongside showing that he is identifiable and that the use was for commercial gain. That is a high bar for a musician with a few thousand listeners. The suit also invoked performers' moral rights under Section 38-B of the Copyright Act, but I have found no Indian ruling applying them to an AI clone of a lesser-known singer.

So, in practice, platform rules may protect a smaller artist faster than a court can. Spotify announced on 25 September 2025 that "vocal impersonation is only allowed in music on Spotify when the impersonated artist has authorized the usage", and said it was putting more resources into its content mismatch process, including letting artists report a mismatch before release.

The new disclosure rules

Since 20 February 2026, the amended IT Rules have required intermediaries that offer tools to make synthetically generated content to label it. MeitY's own FAQ explains that audio needs "a prominently prefixed audio disclosure" and, where technically feasible, permanent metadata identifying it as synthetic. Significant social media platforms must also ask uploaders to declare whether a post is AI-generated before it goes live.

YouTube asks creators to disclose realistic AI-generated or altered content, lists AI-generated music among its examples, and says repeat non-disclosure can lead to removal or suspension from the YouTube Partner Program. Spotify's approach is softer. In April 2026 it launched a beta that shows AI credits for vocals, lyrics or production in Song Credits on mobile, where the artist has disclosed through a label or distributor. Spotify itself says "the absence of a credit doesn't mean AI wasn't used."

What we would do now

If you use AI tools, use them for the parts you are happy to disclose and say so through your distributor if it offers the option. Keep the human parts of the song documented. Put ISRCs on every release, register your compositions with IPRS, and make sure your credits are complete, because any future royalty scheme will run on that data. If you find a clone of your voice, report it to the platform first and keep screenshots and links. And if you are signing anything that mentions AI, voice models or training rights, get a lawyer to read it. Contracts signed now may decide more than the courts do.

We will update this guide when the Delhi High Court rules on the appeal or DPIIT publishes Part 2. For more on what you own and how it pays, read our pieces on who owns Indian indie music masters and Indian indie streaming income.

FAQ

Is AI-generated music copyrighted in India?

The Copyright Act, 1957 says the author of a computer-generated musical work is the person who causes it to be created, but no Indian court has yet applied that clause to a song made with generative AI. The government committee's paper on AI output and ownership had not been published as of October 2026.

Can AI companies use my songs for training without permission in India?

It is not settled. In July 2026 the Delhi High Court held, at the interim stage of ANI's case against OpenAI, that storing works for training appeared to be fair dealing. ANI has appealed. A December 2025 DPIIT proposal would allow training under a mandatory licence with royalties, but it is not law.

What can I do if someone clones my voice with AI?

Report it to the platform. Spotify allows vocal impersonation only when the artist has authorised it. In court, the Bombay High Court protected Arijit Singh's voice on personality rights in 2024, though such claims depend partly on celebrity status.

Do I have to label AI music in India?

The amended IT Rules, in force since 20 February 2026, place labelling duties on platforms and tools, including a spoken disclosure at the start of synthetic audio. YouTube requires creators to disclose realistic AI content, and Spotify offers voluntary AI credits through some distributors.

Will Indian musicians get paid for AI training?

Only if the DPIIT hybrid model or something like it becomes law. The proposal would pay registered rights holders through a central body called CRCAT, using identifiers such as ISRC, ISWC and IPI numbers.