AI Training Data Copyright India: ANI v OpenAI and DPIIT Plan

The law on AI training data copyright India currently applies is the Copyright Act, 1957, which has not been amended for AI. Storing a work electronically is “reproduction” under s.14, but fair dealing for private or personal use including research is allowed by s.52(1)(a). On 24 July 2026 the Delhi High Court refused ANI an interim injunction against OpenAI, taking the prima facie view that storage and training fell within that exception. The suit is undecided, ANI’s appeal is listed for 8 December 2026, and a DPIIT committee has proposed a mandatory licence with royalties that is not yet law.

AI training data copyright India: the basic rules

• What is protected: original literary, dramatic, musical and artistic works, films and sound recordings (s.13). Facts and ideas are not protected; their expression is.
• Originality: needs a minimal degree of creativity and skill, not mere labour (Eastern Book Company v. D.B. Modak, Supreme Court, 2008).
• Exclusive rights: include reproducing the work, which for literary and artistic works covers storing it electronically (s.14).
• Infringement: doing a reserved act without a licence (s.51), unless a s.52 exception applies.
• Fair dealing: s.52(1)(a) permits fair dealing with a work, other than a computer programme, for private or personal use including research, for criticism or review, and for reporting current events.

In every AI training data copyright India dispute, the core fact is that training copies vast amounts of text and images into storage. The question is whether that is infringement or fair dealing.

ANI v OpenAI: what the Delhi High Court held

The news agency ANI sued OpenAI in the Delhi High Court, with summons issued in November 2024. ANI alleged that its news content was used to train ChatGPT without permission, that the model reproduced its content, and that it attributed false statements to ANI. OpenAI argued that facts are not protected, that its use is transformative, that its models are not trained in India, and that publishers can ask to be excluded from future training.

On 24 July 2026 a single judge dismissed ANI’s interim injunction application. As reported:

1. the court accepted territorial jurisdiction, in ANI’s favour;
2. it took the prima facie view that storing ANI’s works fell within s.52(1)(a) and was not infringement;
3. the outputs placed before the court were not substantially similar to ANI’s works, and memorisation or regurgitation was not shown;
4. an injunction would cause irreparable harm to OpenAI and to the public.

Two cautions matter. These are prima facie findings at the interim stage; the suit itself is undecided and ANI’s appeal is pending before a division bench. And the result turned on the evidence of similarity. A case with proof of verbatim reproduction could go the other way.

The DPIIT working paper: a licence-and-royalty model

On 8 December 2025 a committee set up by the Department for Promotion of Industry and Internal Trade published Part 1 of a working paper on generative AI and copyright. Its “hybrid model”:

• a mandatory blanket licence letting developers train on lawfully accessed works without negotiating with each owner;
• royalties payable when the models are commercialised, at rates set by a government-constituted body, open to judicial review;
• central collection and distribution of royalties.

The committee rejected a broad text and data mining exception with an opt-out, which MeitY’s India AI Governance Guidelines had suggested considering, and recorded a dissent from Nasscom favouring such an exception. This is a proposal only. The Copyright Act has not been amended.

What businesses using AI should do now

1. Review outputs before commercial use. Whatever the training position, an output that closely copies an artwork, article, logo, character or song can infringe when you publish it.
2. Upload only what you have rights to. Feeding a competitor’s report, a paid database or a textbook into a tool to “rewrite” it can itself be reproduction, and may breach subscription terms.
3. Add human creative work to important outputs. Ownership of mostly machine-made output is uncertain; for a logo or campaign, keep records of human contribution and register trade marks where relevant.
4. Fix your agency contracts. Under s.17, employees’ works in the course of employment generally belong to the employer, but agencies and freelancers need a written assignment. Add AI disclosure, a similarity check and an indemnity.
5. Read the tool’s terms on output ownership and any IP indemnity.

Example: a Delhi D2C brand’s agency delivers festive packaging made with an image generator. Its central motif closely resembles a known illustrator’s work, and a customer spots it online. The brand withdraws the packaging, settles, and rewrites the agency contract.

Frequently asked questions

AI training data copyright India: is the law settled?

No. The only major ruling, ANI v OpenAI (Delhi High Court, 24 July 2026), refused an interim injunction on prima facie grounds, and the appeal is listed for 8 December 2026. The Copyright Act, 1957 has not been amended for AI training.

Can AI companies train on copyrighted content in India?

The Delhi High Court took the prima facie view in July 2026 that OpenAI’s storage and training fell within fair dealing for private or personal use including research under s.52(1)(a). That is an interim view, not a final ruling, and could change at trial or on appeal.

What did the DPIIT working paper propose?

Published on 8 December 2025, it proposed a mandatory blanket licence for AI training on lawfully accessed works, with royalties on commercialisation set by a government-constituted body. It rejected a text and data mining exception with an opt-out. It is not law.

Who owns AI-generated content in India?

The position is unsettled. Section 2(d)(vi) treats the person who causes a computer-generated work to be created as its author, but works with little human creativity may fail the originality test. Add human creative input and keep records for anything commercially important.

Quick quiz

Protect your business with AI Law and Governance in India, ₹599. Creators can add Law for Content Creators and Influencers, ₹399.

Sources

• Copyright Act, 1957, ss.2(d)(vi), 13, 14, 17, 51 and 52(1)(a)
• ANI Media Pvt. Ltd. v. OpenAI, Delhi High Court, order of 24 July 2026; appeal, notice issued 15 September 2026
• Eastern Book Company v. D.B. Modak, Supreme Court, 2008
• DPIIT committee, Working Paper on Generative AI and Copyright, Part 1, 8 December 2025
• MeitY, India AI Governance Guidelines, 5 November 2025

This page is for education and general information, not legal advice for a specific case. Law as on 31 October 2026.