ANI v OpenAI: Can an AI Company Train Its Model on Copyrighted Content in India?

By Taniya, University Institute of Legal Studies, Himachal Pradesh University, Shimla

Short answer: the question is not yet settled, but for now the Delhi High Court has leaned in the AI company’s favour. On 24 July 2026, it declined to stop OpenAI from storing and using ANI’s news content to train ChatGPT, holding at a prima facie stage that this fell within fair dealing. The order is only an interim one, and the Government has meanwhile proposed a very different fix: a compulsory licence with payment to creators.

This article walks through the dispute, the law behind it, and what law students and exam aspirants should take away from it. It explains the law; it is not legal advice.

1. What is the dispute about?

In November 2024, the news agency Asian News International (ANI) sued OpenAI before the Delhi High Court (ANI Media Pvt. Ltd. v. OpenAI, CS(COMM) 1028/2024) [1]. ANI raised two main complaints:

• The training claim: OpenAI copied and stored ANI’s news content, without a licence, to train ChatGPT.
• The output claim: ChatGPT’s responses reproduced ANI’s content.

ANI sought an interim injunction restraining OpenAI until the case was finally decided. OpenAI argued, among other things, that the Delhi court lacked jurisdiction and that its use of the content was lawful [2].

2. What did the Delhi High Court decide?

On 24 July 2026, Justice Amit Bansal passed a detailed interim order dismissing ANI’s injunction application [2,3]. Three findings stand out.

Jurisdiction: the court held it could hear the case because ANI is based in Delhi and ChatGPT is accessed and used in India, relying on earlier authority that strict territorial limits cannot be applied rigidly to cloud-hosted data [3].

Training: at the prima facie stage, the court held that OpenAI’s storage of ANI’s works for training purposes fell within section 52(1)(a) of the Copyright Act, 1957, the fair dealing provision [2,4].

Outputs: the court reportedly found that ANI had not shown that ChatGPT memorised and reproduced its content [2,4].

These are interim findings only. Courts at this stage look only at whether there is a prima facie case, a balance of convenience and irreparable harm; the judge made clear that none of this binds the court at the final trial, and reports suggest an appeal is pending, so the position could still change [2,4].

3. The law: sections 51 and 52 of the Copyright Act, 1957

Copyright law starts from a simple rule. Under section 51, copying a protected work without the owner’s permission is infringement unless an exception applies [5].

The exceptions sit in section 52, and the one that matters here is section 52(1)(a), which permits fair dealing with a work (other than a computer programme) for:

• private or personal use, including research
• criticism or review
• reporting current events [5]

Notably, the Act has no dedicated exception for text and data mining, the large-scale copying and analysis that AI training involves, unlike the European Union, Japan and Singapore, which have such provisions in some form [6]. In India, the question has to be squeezed into the existing fair dealing categories, usually “private or personal use, including research.” Whether commercial AI training genuinely fits that description remains the real legal controversy, and the interim order is only a first word on it.

Two older Supreme Court decisions are useful background. In R.G. Anand v. Deluxe Films (1978), the Court explained that copyright protects the expression of an idea, not the idea itself [7]. In Eastern Book Company v. D.B. Modak (2008), it held that a work needs a “modicum of creativity” to be protected, rejecting both the “sweat of the brow” approach and an unduly high originality threshold [8]. Both cases are useful when asking exactly what an AI model “takes” from a news article: facts, ideas, or protected expression.

4. Fair dealing versus fair use

These two terms are often confused, but they are different.

• India follows fair dealing: the Act lists specific permitted purposes, and a use must fall within one of them.
• The United States follows fair use: section 107 of its Copyright Act is open-ended and asks a court to weigh four factors, including the effect of the use on the market for the original [9].

The US debate has just moved forward. On 29 September 2026, the Third Circuit became the first US federal appeals court to rule on fair use in AI training, in Thomson Reuters v. ROSS Intelligence. It held that using Westlaw headnotes to train a competing, non-generative legal research tool was not fair use; importantly, it did not decide whether training a generative AI model is fair use [10].

The two rulings are not in conflict. The US case involved a rival product substituting for Westlaw itself; the Delhi case concerns a chatbot, and the Indian court’s finding is interim. Still, the contrast shows how differently the same underlying question can be approached.

5. The Government’s proposal: a compulsory licence

While the courts work through the dispute, the Government has been examining the policy angle. On 8 December 2025, a committee set up by the Department for Promotion of Industry and Internal Trade (DPIIT) released Part 1 of a Working Paper on Generative AI and Copyright [11].

The committee proposed a hybrid model: a mandatory blanket licence for AI training, paired with a statutory right for copyright owners to receive payment, with royalties collected through a collective body. Stakeholder comments were invited until 6 February 2026 [11,12].

The underlying logic is that a free-for-all exception would leave creators unpaid, while requiring permission from every rights owner would make AI development impractical, particularly for Indian start-ups. A statutory licence is offered as the middle path.

Notice the tension with the Delhi order: the DPIIT paper treats unauthorised training as potentially infringing, while the High Court has, at least prima facie, accepted a fair dealing defence [11,12]. If courts eventually hold that training is fair dealing, a compulsory payment scheme would become a fresh policy choice for Parliament to make by amending the Act, rather than a response to settled law.

6. Why this matters for your exams

This is exactly the kind of recent development examiners favour for essay questions, current-affairs sections and mains answers. Keep the following points ready:

• Case name, number and court: ANI Media v. OpenAI, CS(COMM) 1028/2024, Delhi High Court
• Date and nature of the order: 24 July 2026, interim, injunction refused
• Key provisions: sections 51 and 52(1)(a) of the Copyright Act, 1957
• Gap in the law: no express text and data mining exception
• Policy response: the DPIIT hybrid licensing model of December 2025
• Comparison: Indian fair dealing against US fair use, and the Third Circuit’s ROSS ruling

A strong answer will not simply state the outcome; it should weigh both sides. AI developers argue that learning from existing material is how models work, and that heavy restrictions would stifle innovation. Creators and publishers respond that their work is being used to build commercial products without payment, and that the outputs can end up competing with the very material that trained them.

7. Conclusion

India has not yet settled whether training an AI model on copyrighted material is lawful. The Delhi High Court’s interim order currently favours the AI company, but it is not the last word: the final trial, a possible appeal, and the Government’s licensing proposal will all shape the eventual answer. For students, the lesson is to read the order for what it actually decided, remember that interim findings are provisional, and understand the competing interests pulling on both sides.

Sources

1. ANI Media Pvt. Ltd. v. OpenAI Inc., CS(COMM) 1028/2024, Delhi High Court (suit filed November 2024).
2. ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, CS(COMM) 1028/2024, Delhi High Court, interim order dated 24 July 2026.
3. Asia Law Portal, “AI training and copyright” (2026).
4. LexCounsel Law Offices, “India’s First AI Copyright Ruling: The Delhi High Court in ANI v. OpenAI”, Mondaq (12 August 2026).
5. Copyright Act, 1957 (India), sections 51 and 52(1)(a).
6. Copyright Act (Japan), Art. 30-4; Directive (EU) 2019/790, Arts. 3-4; Copyright Act 2021 (Singapore), computational data analysis exception.
7. R.G. Anand v. Deluxe Films, (1978) 4 SCC 118 (Supreme Court of India).
8. Eastern Book Company v. D.B. Modak, (2008) 1 SCC 1 (Supreme Court of India).
9. Copyright Act of 1976 (United States), 17 U.S.C. Section 107.
10. Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., United States Court of Appeals for the Third Circuit, 29 September 2026.
11. Department for Promotion of Industry and Internal Trade (DPIIT), Government of India, Working Paper on Generative AI and Copyright, Part 1 (8 December 2025).
12. Ikigai Law, “Exploring the DPIIT’s Working Paper on Generative AI and Copyright”, Mondaq (15 December 2025).

About the author: Taniya is a law student at the University Institute of Legal Studies, Himachal Pradesh University, Shimla.