The Delhi High Court, in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, has delivered India’s first significant ruling on the application of copyright law to the training and functioning of large language models (“LLMs“).

ANI Media Pvt. Ltd. (“ANI”) is one of India’s most prominent news agencies providing news content to television channels, newspapers and digital platforms across India and abroad. Open AI OpCo LLC (“OpenAI”) is the US-based artificial intelligence company behind ChatGPT, a generative AI chatbot built on LLMs trained on vast datasets, including publicly available web content.

In a suit before Delhi High Court, ANI alleged that OpenAI had infringed its copyright by scraping and storing its news content to train ChatGPT (“training claim”) and by reproducing its copyrighted works in ChatGPT’s responses (“output claim”). In the suit, ANI prayed for various reliefs, including permanent injunction against OpenAI from using its copyrighted news content and grant of damages. Along with the suit, ANI also filed an application seeking interim injunction to stop OpenAI from using its copyrighted content during the pendency of the suit.

The Delhi High Court vide a detailed order dated July 24, 2026, decided and dismissed ANI’s interim injunction application.

On the output claim, the Court held that copyright protects only the original expression of a work and not the underlying facts. It also found that the illustrative ANI articles relied upon in the plaint had been published after the completion of the training of OpenAI’s LLMs and, therefore, could not have formed part of the training dataset. Accordingly, the Court held that the responses elicited through ANI’s adversarial prompts were likely generated through ChatGPT’s Retrieval-Augmented Generation (“RAG“) functionality rather than from memorised training data. Accordingly, ANI failed to establish a prima facie case of copyright infringement in relation to the output claim.

On the training claim, the Court held that while the storage of copyrighted works for training LLMs falls within the scope of the Copyright Act, such use must be assessed together with the defence of “fair dealing” under Section 52(1)(a). It found that the training of LLMs is an internal research process, that OpenAI’s use was limited to training its models, did not substitute ANI’s works or demonstrably affect its market share and served the broader public interest in technological innovation. The Court also held that the balance of convenience favoured OpenAI, noting that ANI had not exercised available opt-out mechanisms to prevent web crawling and that an interim injunction would adversely affect AI development and the wider public interest. Accordingly, the Court declined to grant interim relief.

Prior to dismissing the injunction application, the Court also considered whether Indian courts had jurisdiction despite OpenAI’s servers being located outside India. The Court held that it had the territorial jurisdiction as ANI’s principal place of business is in Delhi and OpenAI offers its services to users in India. It rejected OpenAI’s contention that the location of its servers excluded the application of Indian copyright law, observing that the training activity could not be viewed in isolation merely because the data was ultimately stored overseas where the process involved accessing copyrighted material from India and generating outputs accessible to Indian users.

While disposing of the interim application, the Delhi High Court observed, as is customary in India, that its findings at this stage would not affect the merits of the suit, which will be decided only after a detailed trial. Still, the order is likely to shape AI copyright jurisprudence in India, particularly in light of the DPIIT’s Working Paper proposing a mandatory licensing framework for AI training. The order also signals a judicial inclination towards balancing copyright holders’ rights with the public interest in technological innovation and re-establishes the progressive stance that the Indian judiciary regularly adopts.

Feedback