The Delhi High Court has refused to grant an interim injunction against OpenAI in a copyright case brought by Indian news agency ANI, finding that the use of copyrighted material to train large language models (LLMs) falls within the fair dealing provisions of India's Copyright Act.
Justice Amit Bansal delivered the order on 24 July, holding that OpenAI's storage and use of ANI's news content to train the models underlying ChatGPT is protected under Section 52(1)(a) of the Copyright Act — the provision covering private, personal use and research — and therefore does not constitute infringement.
The court also rejected ANI's argument that ChatGPT reproduces its copyrighted material, finding that the outputs generated by the model were not substantially similar to ANI's original work. Justice Bansal noted that ANI had not established that memorisation or regurgitation of its content had taken place.
Crucially, the judgment observed that it would be economically unviable to develop an LLM if training required licences from multiple sources — a passage likely to be cited widely in future cases. The court added that granting an injunction would cause irreparable harm not only to OpenAI but also to the public at large.
The decision is limited to the interim stage. The case can still proceed to a full trial, and the court acknowledged its observations carry no bearing on the final outcome.
The ruling is the first of its kind by an Indian constitutional court on whether AI companies can use copyrighted news content to train foundational models without a licence. Similar lawsuits involving OpenAI are working through courts in the United States and Canada, making the Delhi decision one that other jurisdictions will watch closely.
For publishers, authors and news organisations in India, the judgment narrows the practical value of pursuing injunctions. Analysts have noted it puts the burden on content owners to prove memorisation rather than mere ingestion — a high evidentiary bar that makes a negotiated licensing deal a more pragmatic path to being paid for archived work.
