IP Law Daily, COPYRIGHT NEWS—High Court of Delhi: AI training found to be fair use, (Jul 27, 2026)

By Matthew Hersh, J.D.

The preliminary ruling finds that the public benefit of large language models outweighed the interests of copyright owners.

An Indian news organization was not entitled to a preliminary injunction against the company behind ChatGPT because the reproduction and storage of copyrighted material to train the artificial intelligence service was consistent with principles of fair use, an Indian court has held. The court, in a decision that drew from principles in the United States and several other jurisdictions, found that OpenAI had established a prima facie case that its large language model served an important public benefit and did not harm the market for the media company’s works.

The decision arises in a lawsuit brought by ANI Media Pvt Ltd, the parent company of South Asian news agency Asian News International. The news agency levels two claims against OpenAI. One, a training claim, asserts that OpenAI committed copyright infringement by the copying and storage of ANI works to train ChatGPT. The other, which the court labels a “reproduction or output” claim, concerns the reproduction of ANI’s works in the output produced by ChatGPT.

The news media organization sought an interim injunction against OpenAI, leading to this opinion.

Jurisdiction. The court first found that it had jurisdiction to hear the lawsuit. OpenAI argued that because the reproduction and training took place outside of India and the training data was stored in US-based servers, Indian copyright law should not apply. But the court rejected that view, finding that the storage of the media company’s works on the US servers was “a terminal step in the chain of events which begin from access of copyrighted works from India and transmission of the same abroad.” To be sure, the court noted, the Indian Copyright Act did not have jurisdiction outside the territory of India. “However,” the court noted, “the Copyright Act does not require this Court to sever the chain of events and examine only the last step.” Thus, at least at this stage of the case, OpenAI’s argument was “untenable as it would lead to evasion of Indian copyright law by infringers who shift the terminal link to servers abroad.”

Reproduction or output claim. The court next found that the media company had not established, at least at this stage of the proceedings, its claim based on ChatGPT’s outputs. The media company alleged that OpenAI’s large language model memorizes and then reproduces substantial amounts of its content in response to user queries. But the court rejected that allegation on two grounds. For one, the court found, in all of the examples that the news media submitted, the material supposedly replicated in ChatGPT’s outputs was created after the training of the model in question—meaning that the materials could not have been memorized by the model. Rather, the court found, it was more likely that ChatGPT simply did fresh website searches in response to user prompts. “Relevant data retrieved from the website, which is contextually aligned with the user’s prompt,” the court surmised, “is then generated as responses grounded in the website’s data rather than relying solely on its pre-trained knowledge.” Thus, the court reasoned, responses that are not based on the trained data “cannot be used against OpenAI in respect of claims based on memorisation.”

In any event, the court found, whatever overlap there was between the media company’s original content and ChatGPT responses to individual user inquiries was not sufficiently substantial under Indian copyright law to count as infringing. Under Indian law, as well as US law, copyright in new reporting subsists “only in the form and manner of expression of news and not on the underlying facts.” The examples given by the media company at this stage of the proceedings fell only into the latter category, the court found. For example, the court noted, one of the alleged examples had to do with the statement by a mother of an Indian athlete after his performance in the 2024 Paris Olympics. The media company argued that ChatGPT replicated an ANI news article about that statement in response to a user inquiry. But “the only similarity which emerges” from a characterization of the news article with the ChatGPT response, the court noted, was the actual quote from the athlete’s mother—something that the media company did not own the rights to. On the whole, the court found, “when the articles of ANI are compared with the responses generated by ChatGPT as a whole, there is no substantial similarity.”

Training claim and fair use. The court then turned to the claim that OpenAI infringed on the media company’s copyrights by reproducing its works as part of the training process. The court concluded, at least prima facie, that the training process involved acts falling within the copyright owner’s exclusive rights. But that raised the more important question of the case: was OpenAI’s use of the content protected by law as fair use? Section 52 of the Indian Copyright Act provides a defense to infringement where there has been a “fair dealing” of any work for “private or personal use, including research,” “criticism or review,” or “the reporting of current events and current affairs.” OpenAI claimed protection under the first prong of that test, covering “private or personal use.” The court found that its claim, at least at this stage of the case, had merit.

The court began by laying out what it understood to be the proper standard for review under fair use principles in Indian law. Although Indian courts had sometimes made reference to the standards developed in other jurisdictions such as the US or the United Kingdom, the court noted, there was “no single test which has been adopted uniformly or consistently by the Indian Courts.” Based on its reading of Indian caselaw, however, the court found that the most important questions were (1) whether OpenAI’s use of the media company’s work was limited to training its LLMs underlying ChatGPT, (2) whether OpenAI’s use of the media company’s works “would result in economic competition and would prejudice the legitimate interests” of the company, and (3) whether the functions performed by OpenAI through its ChatGPT model “serve[d] the public interest.” The court found that the answer to all three of these questions was yes.

The first question was straightforward, the court found, as there was no evidence in the record that OpenAI had used the works of the media company for any purposes other than for training. The other two questions required more analysis but also came out the same way. The court began with the economic competition factor. The media company was a news agency in the business of collecting and distributing news, the court found, that “earns by publishing the news articles on its website, YouTube channel and other social media accounts.” On the other hand, the court noted, OpenAI’s ChatGPT is an application that serves as “tools for content creation, research, language translation, learning and education, summarising, paraphrasing, creation of images based on text description.” In sum, the court noted, “the purpose and character of use of copyrighted works by OpenAI is fundamentally different from that of the news articles of ANI. Hence, it cannot be said that responses produced by ChatGPT are substitutes for news articles published by ANI.”

Moreover, the court found, the public benefits flowing from LLMs underlying ChatGPT were considerable. Those LLMs, the court found, “are designed to assist users in analysing and generating text, improving access to information, supporting education, assisting scientific research, facilitating software development, enabling translation and communication, and creating tools for persons with disabilities.” The technology is also “capable of supporting research, learning, and knowledge dissemination at scale, thereby advancing science and the useful arts in a broader societal sense,” the court emphasized. While the rights of copyright owners remained important, the court noted, “the societal benefits arising from scientific and technological research constitute a relevant consideration in assessing the fairness of a dealing.” Therefore, the court found, “the factor of public interest also stands established.”

MainStory: TopStory Copyright AINews TechnologyInternet GCNNews