What the court actually held
Justice Amit Bansal declined to grant ANI Media an interim injunction against OpenAI, and the reasoning is what matters. The court found that OpenAI's storage and use of ANI's news reports to train its large language model falls prima facie within 'fair dealing' under Section 52(1)(a) of the Copyright Act, 1957, which exempts certain research and private uses. It also held that ANI had not shown, at this stage, that ChatGPT reproduced or retrieved its articles in the answers it generates.
This is a preliminary ruling on whether to freeze OpenAI's conduct while the case proceeds, not the final word on liability. The case drew interventions from the Federation of Indian Publishers, the Digital News Publishers Association and the Indian Music Industry, because the answer sets the tone for every publisher-versus-AI fight in the country. For now the tone favours the model builder.
Why this does not settle it in Europe
An EU owner should not treat an Indian fair-dealing finding as a green light at home. Europe runs on a different rule: the text-and-data-mining exception in Article 4 of the 2019 Copyright Directive lets rightsholders opt out and reserve their works, which is exactly the mechanism news publishers are now invoking. The Court of Justice of the EU is already weighing a Hungarian reference, Like Company v Google Ireland, on whether an LLM's answers reproduce press content and whether training is a reproduction at all. Until that lands, the European position is unsettled, and the UK is running its own separate consultation.
So the same model can rest on training data ruled acceptable in Delhi and contested in Brussels. That gap is the point. A licence a US or Indian vendor did not need may be one a European deployment does, and the cost of getting it wrong, in euros or pounds of damages, sits with whoever put the tool into production, not only with the lab that trained it.
What it changes for you
Treat training-data legality as a contract term, not a settled fact. The Delhi ruling lowers the odds that the specific ANI claim halts ChatGPT, and it gives every AI vendor a favourable precedent to cite, which will show up in sales decks within the week. It does not remove your exposure if you deploy the output in a market where the law bends the other way.
The move worth making this quarter is narrow and concrete. Ask any AI vendor you rely on for a written indemnity covering copyright claims on training data, and read whether it is capped, carved out for open-weight components, or silent on your jurisdiction. A precedent in India is useful; an indemnity that pays your legal bill in Frankfurt or London is the thing you can actually hold.
Read next: A US Bill Could Switch Off the AI You Depend On | Sony Fingerprinted What Was Inside The Model
