The single judge held in July that storing news content to train a large language model prima facie falls within the fair dealing exception, and that an injunction would harm AI development and the public. The news agency says that reads a closed statutory exception far wider than Parliament wrote it. Notice has now issued on the appeal.
New Delhi: The Delhi High Court has issued notice to OpenAI on an appeal by news agency Asian News International against a single judge’s refusal to grant it an interim injunction in its copyright infringement suit over the use of its content to train ChatGPT.
A Division Bench of Justice Avneesh Jhingan and Justice Manmeet Pritam Singh Arora issued notice and listed the matter for hearing on December 5. The Bench declined, at this stage, to issue any direction concerning an earlier arrangement on scraping without first hearing OpenAI’s response.
The order under challenge
The suit was filed in 2024, and is the first by an Indian media house against OpenAI alleging copyright violation. The agency’s case is that its content was used without authorisation to train and operate ChatGPT, and it sought an injunction restraining the storage, publication, reproduction or use of its copyrighted works while the suit is pending.
In July 2026, Justice Amit Bansal refused that relief. On a prima facie view, he held that the storage of the agency’s literary works for the purpose of training the models underlying ChatGPT falls within the fair dealing exception in Section 52(1)(a) of the Copyright Act, 1957, covering private or personal use including research, and therefore does not amount to infringement. The judge also held that irreparable injury would be caused not only to OpenAI but to the public interest if an interim injunction were granted.
Two findings in that order went the agency’s way and are worth recording, because they are often lost in the summaries. The Court held that the agency retains copyright in its original literary works notwithstanding that they are available free of charge on its website, and it rejected OpenAI’s objection to territorial jurisdiction, holding that the Delhi High Court could entertain the dispute despite the servers being located outside India.
The findings were expressly prima facie. The suit continues, and the separate claim concerning hallucinated stories attributed to the agency remains alive.
The grounds of appeal
The appeal attacks the fair dealing reasoning at its foundation. The agency’s position is that India has fair dealing rather than fair use, that Section 52 sets out a closed list of permitted purposes, and that reading commercial exploitation at scale into the words private or personal use, including research, widens the exception beyond what Parliament enacted. The distinction matters: American fair use is an open-ended balancing test, while the Indian provision enumerates specific purposes and admits of no general residual category.
On the public interest finding, the agency argues that it sits uneasily with OpenAI’s own pleading that the agency’s content is a minuscule fraction of its training data. If the material is that marginal, the argument runs, restraining its use cannot plausibly imperil AI development in India.
The third ground is the licensing market. The agency points out that OpenAI licenses news content for training and distribution from publishers including the Financial Times, and contends that if the use were genuinely fair dealing it is hard to see what those licences are for. The existence of that market, it says, is precisely what a fairness analysis is meant to weigh.
It has also argued that making content publicly accessible does not confer a right to exploit it or to make and store copies of it.
The scraping undertaking
Advocate Sidhant Kumar, for the agency, told the Bench that OpenAI had earlier volunteered an interim arrangement not to scrape content from the agency’s website, and that the position had held from September 2024 until the impugned order was passed.
The Bench indicated that it was not inclined to issue any direction on that arrangement without first hearing OpenAI, and confined itself to issuing notice.
The intervenor
Senior Advocate Kapil Sibal, appearing for the Broadband India Forum, pressed to be heard by the Division Bench, noting that the forum had been heard by the single judge and that the matter carries global implications.
The Bench questioned whether the forum was an impleaded party and whether it could be a party at every stage, asking counsel to assist on the next date. Senior Advocate Arvind Datar, also for the forum, said a formal intervention application would be filed.
Why it matters
The July order is the only Indian judicial pronouncement on whether an AI developer may train on copyrighted news, and it is doing work far beyond the two parties. Every licensing negotiation between an Indian publisher and an AI company is now conducted in its shadow, because it sets the value of the alternative to a licence. If training is fair dealing, a publisher’s bargaining position is considerably weaker than it was before July.
That is why the appeal matters more than the ordinary interim appeal. A Division Bench now has the opportunity to test a reading of Section 52(1)(a)(i) that stretches a phrase drafted for individual study and scholarship across industrial-scale ingestion. There is a respectable argument each way. The single judge reasoned that Section 52 expressly requires non-commercial use in some clauses and does not do so in this one, so no automatic commercial-use bar can be implied. The agency’s answer is that the absence of an express bar does not convert a closed enumerated exception into an open-ended one.
The public interest limb is the more unusual part of the order and the most exposed on appeal. Balance of convenience in copyright cases normally weighs the harm to the parties, not the developmental trajectory of an industry. Importing the latter invites the question the agency has now put: if this content is a negligible share of the training corpus, whose development is actually at stake.
There is also a policy current running the other way. A central government panel has taken the view that AI training on copyrighted content should be paid for, a position that does not sit comfortably alongside a judicial reading under which no payment is required at all. Where the courts land will determine whether that policy has anything left to regulate.
For now, nothing has changed. Notice has issued, the single judge’s order stands, and the position on scraping awaits OpenAI’s response in December.
Case: Appeal by ANI Media Private Limited against the order dated July 24, 2026 of a single judge in ANI Media Private Limited v. OpenAI OpCo LLC [CS(COMM) 1028/2024]
Bench: Justice Avneesh Jhingan and Justice Manmeet Pritam Singh Arora, High Court of Delhi at New Delhi
Date of Order: September 15, 2026 | Next Listed: December 5, 2026
Appearances: Advocate Sidhant Kumar for the appellant. Senior Advocates Kapil Sibal and Arvind Datar for the Broadband India Forum, intervenor.
Status: Notice issued. No direction passed on the scraping arrangement. Findings of the single judge are prima facie and the suit remains pending.