Does this ruling mean AI companies can freely train on Indian news content now?

Not exactly. The court only denied ANI’s request for an interim injunction based on a preliminary view of the case. The underlying infringement suit is still pending, and the judge noted his reasoning on Section 52 could be revisited on appeal or at trial.

What is Section 52 of India's Copyright Act?

It lists exceptions to copyright infringement under Indian law, including use for private or personal purposes such as research — the fair-dealing provision the court applied to OpenAI’s training data storage.

Why did the court still rule in ANI's favor on jurisdiction?

Jurisdiction and copyright infringement are separate legal questions. The court found it had authority to hear the case even as it declined, at this stage, to find OpenAI’s conduct infringing.

Delhi High Court Clears OpenAI's ChatGPT in ANI Copyright Suit

The Delhi High Court has declined to stop OpenAI from using news agency ANI’s articles to train ChatGPT, issuing India’s first substantive court ruling on whether AI companies can train large language models on copyrighted journalism without a license.

Justice Amit Bansal handed down the 135-page order on July 24, rejecting ANI’s request for an interim injunction in a suit the agency filed against OpenAI in November 2024. On a preliminary basis, the court found that OpenAI’s storage of ANI’s articles to train its models falls within Section 52(1)(a)(i) of India’s Copyright Act — the fair dealing exception for private or personal use, including research.

No evidence of memorization

A second question was whether ChatGPT itself infringes copyright when it answers questions using ANI’s reporting. The court put the burden on ANI to show the chatbot reproduces its work rather than merely draws on it, and found the agency had not established that ChatGPT “memorised or regurgitated” its articles in the responses it generates for users.

Justice Bansal also weighed the public interest, concluding that an injunction would disrupt access for millions of Indian ChatGPT users and set back domestic AI development — a factor he said outweighed ANI’s claims at this preliminary stage.

The ruling wasn’t a clean sweep for OpenAI: the court sided with ANI on jurisdiction, confirming the Delhi High Court has the authority to hear the case. And the decision covers only ANI’s request for interim relief — the underlying infringement suit continues, and the court noted its reasoning could still be revisited on appeal, since Section 52 had never before been tested against foundation-model training.

The case joins a growing global docket of publisher lawsuits over AI training data, an area of litigation that has produced sharply different outcomes so far — from Anthropic’s $1.5 billion settlement over pirated books to unresolved fights over what counts as fair use of copyrighted text, as our explainer on AI training and fair use covers in more detail.

No evidence of memorization

Frequently asked questions