AI and Copyright — Who Owns AI-Generated Content in India?

Introduction

The advent of AI has revolutionised browsing, research and content creation. These tools can absorb vast troves of data, synthesise it, and generate something novel from it. "But this leap brings with it a critical legal question of the ownership of content created by AI. Is it the AI itself, the person who provided the prompt, or the sources from which the AI has learned?" [1]This blog examines that question through the lens of India's Copyright Act, 1957, a statute that, at its core, assumes every author is a human being, and has yet to reckon with a creator, in this case the AI, whom the law doesn't even recognise as a person.

Section 2(d) of the Copyright Act, 1957 defines "author" as the person who physically creates a work the writer, composer, artist, or photographer with two exceptions: for films and sound recordings, it is the producer, and for computer-generated works, it is "the person who causes the work to be created."[2]The provision itself doesn't explicitly demand a human author, but courts and commentators have consistently read in a requirement of legal personhood with AI or the software acting only as a medium of creativity. “Because generative AI often produces outputs without a clearly identifiable human hand behind the specific creative choices, existing law generally fails to recognise such works as copyrightable at all. Users who supply prompts may claim ownership, but their contribution is often not creative enough to count as authorship in the traditional sense; developers who build the AI tools, meanwhile, cannot claim rights either, since they play no role in generating any specific output.”[3]

This ambiguity isn't just theoretical. The law was written in 1994 at a time when "computer-generated" meant something narrow and rule-based, like commanding a computer to do a certain task or simply follow the instructions of the user. [4] In those cases, a human operator clearly steered the outcome, so there was no dubiousness about who "caused" the work. Generative AI breaks this chain of causation: a prompt nudges a probabilistic model that draws on billions of learned patterns, and the same prompt can produce entirely different outputs each time. The provision presupposes a straight line from human instruction to predictable result; a line generative AI simply does not follow[5]

The Raghav Case: A Crack in the Foundation

The clearest illustration of this confusion played out through RAGHAV, an AI painting tool built by Ankit Sahni. Sahni first registered a piece of AI-assisted art listing only himself as author, which the Copyright Office accepted without issue. He then filed a second application listing himself and RAGHAV as co-authors and remarkably, this was accepted too, making it one of the first instances globally of a copyright office recognising an AI as a co-author.[6]

The Copyright Office subsequently issued a notice seeking to withdraw that registration, prompting Sahni to challenge the withdrawal through due process. The matter remains unresolved, with the registration status still listed as "registered" even as the underlying legal question sits unanswered.[7]The episode exposed a real gap: India's own copyright bureaucracy could not arrive at a consistent position on AI co-authorship, let alone a court. Legal scholars have since proposed splitting AI-generated works into two categories "AI-assisted," where substantial human creative input exists and authorship should vest in the human user, and "AI-autonomous," where human input is minimal and the work might instead be pushed into the public domain or governed by a new sui generis right.[8]

The Bigger Fight: ANI v. OpenAI

While Raghav concerned authorship of AI outputs, a larger battle is unfolding over what goes into AI models in the first place. Since November 2024, the Delhi High Court has been hearing ANI Media Pvt Ltd v. OpenAI Inc., India's first major lawsuit testing whether training a large language model on copyrighted material amounts to infringement.[9] ANI alleges that OpenAI scraped its news articles, including paywalled content, to train ChatGPT, and that the chatbot has, at times, reproduced its reporting or fabricated false attributions to it.[10] OpenAI's defence rests on three pillars: that its servers are not located in India, raising jurisdictional questions; that training extracts statistical patterns rather than "expressive" content; and that any temporary storage during training does not amount to reproduction under copyright law.[11] ANI counters that Section 52's "fair dealing" exceptions are exhaustive and do not stretch to commercial AI training, and that infringement occurs at the point copyrighted material is scraped and stored — not only if it later resurfaces in an output.[12] The case has been drawn in the wider industry as a proxy war. The Digital News Publishers Association, Indian Music Industry, and Federation of Indian Publishers have intervened supporting ANI, while the Broadband India Forum and several AI startups have intervened backing OpenAI, warning that strict liability could stall India's AI sector.[13]After more than 32 hearings, Justice Amit Bansal reserved the Court's order on interim relief in early 2026 — meaning a ruling could arrive at any time, and whichever way it goes, it will be the first judicial word on how Indian copyright law treats AI training data.[14]

How the UK Does It Differently

India's Section 2(d)(vi) closely mirrors Section 9(3) of the UK's Copyright, Designs and Patents Act, 1988, which assigns authorship of a computer-generated work to "the person by whom the arrangements necessary for the creation of the work are undertaken."[15]Notably, the UK provision goes a step further than India's by explicitly defining computer-generated works as those made "in circumstances such that there is no human author" directly contemplating fully autonomous creation in a way India's provision never quite addresses.[16] Even this clearer language hasn't fully solved the UK's own problem, and its government has since run consultations on reforming copyright specifically for AI, suggesting India isn't alone in playing catch-up; it simply started the race further behind.

A way forward

The good news is that India doesn't need to invent a solution from scratch  it needs to choose between paths other jurisdictions are already testing, and adapt one to its own statute. The first option is a legislative amendment to Section 2(d)(vi) itself, borrowing the UK's precision. Explicitly defining "computer-generated work" as one made "in circumstances where there is no human author," and clarifying that the "person who causes the work to be created" can include someone who exercises meaningful creative control through prompting, curation, or editing would close the interpretive gap that RAGHAV fell into. This doesn't require treating AI as an author; it requires being honest about who the human author actually is when the process is iterative rather than instantaneous.

The second is the two-tier framework legal scholars have already proposed: splitting outputs into "AI-assisted" works, where authorship vests in the human, and "AI-autonomous" works, where human input is negligible and the output either falls into the public domain or is governed by a new, narrower sui generis right, similar to the protections some countries extend to databases or broadcasts. This avoids overreaching by handing AI-autonomous content the same 60-year copyright term as a human-authored novel, while still giving genuine human effort a home in the existing law.

Third, and most immediate: the Delhi High Court's ruling in ANI v. OpenAI will not settle authorship questions, but it will settle the upstream question of whether training itself is lawful, and that ruling will shape every downstream case, including future RAGHAV-style disputes. A finding that scraping and storage constitute infringement absent a licence would push Indian AI companies toward licensing deals with publishers, the way several US and European outlets have already struck with OpenAI and other labs, effectively creating a market solution alongside whatever the statute eventually says.

In the interim, the practical advice for anyone building on AI tools in India isn't to wait for Parliament. It's to document the human contribution as it happens: save prompt iterations, editing history, and curation decisions, since that paper trail is what will separate a defensible authorship claim from a one-line prompt gamble if and when a court is asked to draw the line. Until the Commerce Ministry's panel reports back, or the Delhi High Court rules, that documentation is the only real insurance available.

[1] Tanishka Sinha, Ownership of AI Generated Content: A Deep Dive into Copyright Law in India, Khurana & Khurana (Oct. 25, 2025), https://www.khuranaandkhurana.com/ownership-of-ai-generated-content-a-deep-dive-into-copyright-law-in-india.

[2] The Copyright Act, No. 14 of 1957, § 2(d), India Code (1957).

[3] Ketan Joshi, Stealing Ideas: Intersection of AI-Generated Art and Copyright Law, Lexology (Maheshwari & Co. June 20, 2025), https://www.lexology.com/library/detail.aspx?g=0792eed4-bb5d-4aaa-aea8-498b97875e6c.

[4] Noel Naiju George, Adapting Indian Copyright Law to the Age of Artificial Intelligence: Recognizing AI as Authors Under the Copyright Act of 1957, 4 Indian J. Integrated Rsch. L. 810, 810 (2024).

[5] Shashank Tripathi, Navigating Indian Copyright Framework in the Age of AI Generated Works, NLIU Cell for Stud. Intell. Prop. Rts. (Oct. 10, 2023), https://csipr.nliu.ac.in/miscellaneous/navigating-indian-copyright-framework-in-the-age-of-ai-generated-works/

[6] Sinha, supra note 1.

[7] Id.

[8] Madhuri Hardia & Navin Dave, Copyright Protection for AI-Generated Works in India, 5 Int'l J. Progressive Rsch. Engineering Mgmt. & Sci. 822, 822 (2025).

[9] OpenAI Faces Data Scraping Allegations in India's First-Ever Generative-AI Copyright Infringement Suit, World Trademark Rev. (Apr. 17, 2025), https://www.lexology.com/library/detail.aspx?g=15927c99-bfb6-4ed3-8671-f85ec42217b2.

[10] Jurisdiction of Delhi High Court and Copyright Implications in AI Training: ANI v. OpenAI Case, The Legal Aff. (2025), https://thelegalaffair.com/news/jurisdiction-of-delhi-high-court-and-copyright-implications-in-ai-training-ani-v-openai-case/.

[11] Rimjhim Singh, ANI-OpenAI Dispute: Delhi High Court Reserves Order in AI Copyright Case, Bus. Standard (Apr. 2, 2026), https://www.business-standard.com/companies/news/ani-openai-case-delhi-high-court-ai-copyright-dispute-chatgpt-training-data-126040200589_1.html.

[12] The Future of Indian Copyright Legislation in the Wake of ANI Media v. OpenAI, IAM (Nov. 25, 2025), https://www.iam-media.com/guide/india-managing-the-ip-lifecycle/2026/article/the-future-of-indian-copyright-legislation-in-the-wake-of-ani-media-v-openai.

[13] Delhi HC Reserves Verdict in ANI-OpenAI Copyright Case, Pitchonnet (Apr. 2, 2026), https://www.pitchonnet.com/pitch-feature/delhi-hc-reserves-verdict-in-ani-openai-copyright-case-39529.html.

[14] Samana, Generative AI vs. Copyright Law: Why Existing Legal Frameworks Are Failing Creative Professionals, Prime Legal Blog (May 9, 2026), https://blog.primelegal.in/india-first-ai-copyright-case-ani-vs-openai-delhi-high-court/.

[15] Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (UK).

[16] Hardia & Dave, supra note 8.