Who owns the machine's imagination? AI, copyright and the Indian courts

Artificial Intelligence has transformed the way creative works are produced, challenging traditional ideas of authorship, originality, and ownership. Today, AI systems can generate text, music, images, software, and other forms of creative expression with minimal human intervention. This technological development raises a fundamental legal question: when a machine creates, who owns its imagination?

Copyright law has traditionally been built around the idea of a human creator who exercises creativity and produces an original work. However, AI-generated works complicate this framework by separating the act of creation from the conventional human author. Questions arise as to whether an AI-generated work can receive copyright protection, whether the person who operates or trains the AI can claim ownership, and how much human involvement is necessary for copyright to arise.

The Copyright Act, 1957 was never written with machines in mind, yet it contains a clause that has suddenly become the most litigated provision in Indian IP law. Section 2(d)(vi) defines the "author" of a computer-generated literary, dramatic, musical or artistic work as "the person who causes the work to be created." That single phrase - "causes the work to be created" - is where every current Indian dispute over AI authorship begins.

One of the most famous test cases is the RAGHAV matter. In this case, the Indian Copyright Office briefly registered an AI tool, RAGHAV, as co-author alongside its human user, Ankit Sahni, for an AI-assisted artwork. Following scrutiny, the Office withdrew the AI's co-author status, leaving the human prompter as sole author - signalling that Indian law is prepared to protect AI-assisted work, but only where a human has exercised enough "skill and judgment" to be treated as the true creative cause.

Indian copyright law may protect work created with the assistance of AI, but the protection depends on meaningful human involvement. If a person uses AI merely as a tool while contributing sufficient skill, judgment, creativity, or intellectual effort, the human may be recognized as the true author of the resulting work. However, where the AI independently generates the work with little or no meaningful human contribution, claiming copyright protection becomes much more difficult. A generic prompt such as "draw a landscape" is unlikely to meet that bar; a carefully directed, iteratively refined creative process is more likely to. The line between the two remains for courts to draw, case by case.

That unfinished business is now before the courts directly. In Stephen Thaler v. Union of India, concerning a four-year-pending application to register a purely AI-generated artwork with no human author, the matter came up again this April, with the court directing the Copyright Office to actually progress the long-stalled hearing rather than let it languish. India's answer to the harder question - whether a work with no human author at all can be copyrighted here - is still awaited.

The bigger and more consequential battle, however, is not about who owns AI output, but about what AI is permitted to consume. In ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, India's leading news agency accused OpenAI of training ChatGPT on its copyrighted reportage without licence, and of generating fabricated content falsely attributed to ANI. In an interim order delivered on 24 July 2026, Justice Amit Bansal of the Delhi High Court made two findings that will reverberate well beyond this one case. First, he rejected OpenAI's argument that Indian courts lack jurisdiction merely because its servers sit abroad, holding that Indian courts can hear such disputes wherever the effects of the alleged conduct are felt in India. Second, on a prima facie, interim-stage view, the Court found that storing ANI's content to train the model fell within the fair dealing exception under Section 52(1)(a) of the Act, and that ChatGPT's retrieval-augmented outputs were not substantially similar to ANI's original reports. Importantly, this is not a final verdict, nor a blanket licence for all AI training in India - the underlying suit continues on separate claims, and Section 57 of the Act still allows authors or their representatives to object to prejudicial distortion of their work independently of who holds the economic copyright.

A parallel front has opened around personality rights, where Indian courts have moved faster and more decisively than on training-data copyright. Bollywood playback singer Arijit Singh secured protection from the Bombay High Court against AI tools cloning his voice; Asha Bhosle obtained similar ad-interim relief against AI companies replicating her voice and likeness; and Aishwarya Rai Bachchan has since moved the Delhi High Court over unauthorised AI use of her image. These cases build on earlier precedent protecting the persona of figures like Anil Kapoor and Amitabh Bachchan, and confirm that even where Indian copyright law is still finding its footing on AI, the common-law right to one's own face, voice and identity is already being enforced against AI misuse - often more readily than copyright claims themselves.

For public, three implications follow. First, "who is the author" and "was the training lawful" are now two separate, live questions in Indian law, and clients need advice on both - not a general answer that AI content is or isn't protected. Second, jurisdiction is no longer a reliable shield for foreign AI companies; Indian claimants can sue where the harm lands, not only where the servers sit. Third, until Parliament legislates specifically on text-and-data-mining exceptions - as Japan and the EU already have - Indian courts will keep deciding these disputes through the imperfect fit of Section 52's fair dealing exception, on a case-by-case, often interim basis.

None of these rulings is final. The ANI suit continues to trial. The Thaler matter awaits a substantive hearing. What is settled, for now, is only this: Indian courts are willing to hear these disputes and willing to assert jurisdiction over global AI platforms. The law is being written, quite literally, one interim order at a time.

The author is a practising advocate of Patna High Court, and specialises in international laws, corporate laws and cyber laws; Views presented are personal.