Delhi High Court Lays Down India’s First Judicial Framework on Copyright and Generative AI

In ANI Media Pvt. Ltd. v. Open AI OpCo LLC, CS(COMM) 1028/2024, the Delhi High Court delivered a landmark judgment on the intersection of copyright law and Generative Artificial Intelligence. The ruling addresses AI training, AI-generated outputs, fair dealing, and the jurisdiction of Indian courts over overseas AI developers.

Summary

Deciding an application for interim injunction, the Court examined questions concerning the use of copyrighted works for training Large Language Models (LLMs), the scope of copyright infringement arising from AI-generated outputs, the applicability of the fair dealing exception under Section 52 of the Copyright Act, 1957, and the territorial jurisdiction of Indian courts over AI developers operating through servers located overseas.

While refusing interim injunctive relief against OpenAI, the Court examined these issues through established principles of Indian copyright law. The Court held, on a prima facie basis, that OpenAI’s storage of ANI’s copyrighted works for training its LLMs fell within the scope of Section 52(1)(a) of the Copyright Act and therefore did not amount to infringement under Section 51.

The Court also found, prima facie, that the outputs relied upon by ANI did not substantially reproduce its copyrighted works.

Importantly, the judgment was confined to the adjudication of ANI’s application for interim relief. The Court clarified that its observations would have no bearing on the final outcome of the suit.

Facts of the Case

ANI Media Pvt. Ltd. instituted a copyright infringement suit alleging that OpenAI had used its copyrighted news articles without authorisation for training the Large Language Models underlying ChatGPT.

ANI’s Claims

ANI advanced two principal claims:

1. Training claim — OpenAI copied and stored ANI’s literary works on foreign servers for training its AI models.
2. Output claim — ChatGPT reproduced or substantially reproduced ANI’s copyrighted works while responding to user prompts.

ANI argued that both activities infringed its exclusive rights under Sections 14 and 51 of the Copyright Act, 1957.

OpenAI’s Defence

OpenAI disputed these allegations, contending that LLMs do not ordinarily retain or reproduce copyrighted material in a manner amounting to substantial reproduction, but instead learn statistical relationships and generate responses based on their training. It further argued:

• The relevant training activities occurred on servers located outside India.
• AI-generated outputs did not amount to substantial reproduction of ANI’s expression.
• The storage and use of publicly available material for training could fall within the fair dealing exception under Section 52 of the Copyright Act.

Owing to the novel issues involved, the Court appointed two Amici Curiae and permitted several industry bodies, publishers, AI companies and policy organisations to intervene in the proceedings.

Issues before the Court

The Court considered four principal questions:

1. Whether Indian courts possessed jurisdiction despite OpenAI’s servers being located outside India.
2. Whether storage of ANI’s copyrighted works for training LLMs amounted to copyright infringement.
3. Whether AI-generated outputs constituted substantial reproduction of copyrighted literary works.
4. Whether the use of copyrighted material for AI training could be protected as fair dealing under Section 52(1)(a) of the Copyright Act, 1957.

Findings of the Court

Territorial Jurisdiction

The Delhi High Court held that it possessed territorial jurisdiction under Section 62(2) of the Copyright Act and Section 20 of the Code of Civil Procedure, 1908.

The Court took into account, among other factors, ANI’s presence within its territorial jurisdiction, the availability of OpenAI’s services to users in India, and the generation and communication of the outputs relied upon by ANI to users in India. The Court further held that the location of OpenAI’s servers outside India, by itself, could not defeat the jurisdiction of Indian courts.

Copyright Infringement and Substantial Reproduction

On the issue of copyright infringement, the Court drew an important distinction between AI training and AI-generated outputs. It observed that, on the material presently available, there was insufficient evidence to conclude that ChatGPT ordinarily memorised and reproduced copyrighted works in a manner amounting to substantial reproduction under the Copyright Act.

While acknowledging that “regurgitation” or memorisation may occur in isolated instances, the Court found that ANI had not established such memorisation or regurgitation in the material placed before it. Whether memorisation actually occurred in a particular case and whether any output amounted to substantial reproduction would depend upon the evidence and the circumstances of the individual case.

Fair Dealing under Section 52(1)(a)

The Court then undertook an extensive interpretation of Section 52(1)(a) of the Copyright Act in the context of Generative AI. It adopted a two-stage approach comprising a “Purpose Test” and a “Fairness Test” for determining whether the use of copyrighted works could constitute fair dealing.

The Purpose Test

On the Purpose Test, the Court considered whether the use of copyrighted works for training LLMs could fall within the purposes specified under Section 52(1)(a), including “private or personal use, including research”.

The Court rejected a general proposition that commercial use, by itself, would necessarily exclude the availability of the fair dealing exception. The Court noted that where Parliament intended to impose a non-commercial limitation in Section 52, it had expressly done so.

The Court also considered whether the concept of “research” could be applied to the training of LLMs. Applying the doctrine of updating construction, the Court observed that statutory language may, where appropriate, be interpreted in light of developments in science and technology, provided that such interpretation remains consistent with legislative intent.

On this basis, the Court held, at the prima facie stage, that the training of LLMs could fall within the scope of “research” contemplated under Section 52(1)(a).

The Fairness Test

The Court thereafter considered whether OpenAI’s dealing with the copyrighted works was fair. In doing so, it considered:

• The nature and purpose of the use.
• The interests of copyright owners.
• The potential impact on those interests.
• The broader public interest in technological development and innovation.

Applying these principles, the Court held, on a prima facie basis, that OpenAI’s storage of ANI’s copyrighted works for training the LLMs underlying ChatGPT fell within Section 52(1)(a) and therefore did not amount to infringement under Section 51.

The Court’s conclusion was confined to the interim stage and did not constitute a final determination on the copyrightability or legality of AI training.

Held

Finding that ANI had failed to establish a prima facie case warranting interim injunctive relief, the Delhi High Court dismissed the application for interim injunction.

The Court held that the balance of convenience did not favour the grant of an injunction at this stage, particularly where the questions concerning memorisation, substantial reproduction and copyright infringement required further evidence and detailed examination. The suit was directed to proceed for final adjudication.

The Court also clarified that the observations made in the judgment were confined to the adjudication of the interim injunction application and would have no bearing on the final outcome of the suit.

Conclusion

The judgment marks a watershed moment in Indian copyright jurisprudence by establishing the country’s first judicial framework governing Generative AI. Rather than creating new principles, the Delhi High Court adapted existing copyright doctrines to emerging AI technologies through a structured analysis of jurisdiction, infringement, substantial reproduction and fair dealing.

Its recognition that AI training may qualify for protection under Section 52 while simultaneously preserving copyright remedies against infringing outputs strikes a careful balance between innovation and authors’ rights.

Equally significant is the Court’s refusal to treat the foreign location of AI infrastructure as a jurisdictional bar, ensuring that Indian copyright owners retain effective remedies against global AI platforms operating within India.

The decision is likely to serve as the foundational authority for future disputes involving Generative AI, copyright licensing, AI training datasets and the evolving relationship between intellectual property law and artificial intelligence.

Last Updated on 10 August, 2026