The Delhi High Court’s latest order in the copyright dispute between news agency ANI and OpenAI does not settle whether artificial intelligence companies can freely train models on copyrighted material.
But it does deliver an important interim signal: at least in this case, the court was not persuaded that ANI had established grounds to immediately restrict OpenAI’s use and storage of its content.
That distinction matters. The July 24 ruling is not a blanket licence for AI companies to use copyrighted works. It is an interim decision in a case that remains unresolved, but one that could influence how India’s courts, publishers and technology companies approach the increasingly contentious economics of AI training data.
OpenAI vs ANI
ANI’s lawsuit against OpenAI, filed in November 2024, alleged that the AI company had used its copyrighted material without authorisation to train its models. ANI also alleged that ChatGPT had generated fabricated news reports attributed to the agency.
The Delhi High Court, however, rejected ANI’s request for immediate interim relief. Reuters reported that the court did not find sufficient evidence at this stage that ChatGPT had memorised or reproduced ANI’s original articles in a manner that established a prima facie copyright infringement case.
The distinction between training data and AI-generated output is central to the dispute. Copyright law protects original works, but determining whether the process of training an AI model on such works constitutes infringement is a relatively new legal question.
The court’s interim view therefore provides OpenAI with room to continue its operations while the broader lawsuit proceeds. It does not, however, eliminate the underlying copyright dispute.
Section 52 Becomes Key Issue
One of the most important aspects of the ruling concerns Section 52(1)(a)(i) of India’s Copyright Act, 1957. The provision contains exceptions for fair dealing with copyrighted works for purposes including private or personal use and research.
The court’s interim reasoning considered OpenAI’s storage of ANI’s works within the scope of this provision. That is significant, but the finding needs to be understood narrowly.
The judgment does not establish that every form of AI training involving copyrighted material qualifies as fair dealing. Nor does it mean that AI companies can reproduce or commercially exploit copyrighted works without restriction.
Instead, the case highlights a legal gap that is becoming increasingly important. India’s copyright framework predates the widespread adoption of generative AI and was not designed specifically to address how large language models acquire, store and process vast amounts of information.
For technology companies, the distinction could affect the legal risks associated with model development. For publishers, it raises a different concern: whether their content can contribute to the development of commercially valuable AI systems without a licensing arrangement.
Publishers Face A New Challenge
The dispute reflects a broader economic tension between content creators and AI companies.
News organisations spend money on journalists, editors, photographers and infrastructure to produce original reporting. AI systems, meanwhile, depend on enormous datasets to improve their ability to understand and generate language.
The question is therefore not simply whether content is publicly accessible. It is whether access to that content creates commercial value for an AI company and, if so, whether the original creator should receive compensation.
The Delhi High Court’s interim decision does not resolve that question. But it could affect the bargaining dynamics between publishers and technology companies.
If courts ultimately allow broader use of publicly available content without mandatory licensing, publishers may have to explore alternative ways of protecting their commercial interests. These could include licensing agreements, contractual restrictions, technical controls or negotiations over attribution and access.
The opposite outcome could produce a different market, where AI companies are required to negotiate with large numbers of copyright holders or develop licensing frameworks for high-value datasets.
India Becomes A Legal Test
The ANI case is important because India is becoming a significant market for both digital content and artificial intelligence. The country has a large English-language news ecosystem, a growing digital economy and an expanding user base for generative AI products.
That makes the legal treatment of copyrighted material particularly consequential.
A final judgment in the ANI case could influence how publishers assess their rights and how technology companies structure their data practices in India. It could also become relevant to future disputes involving books, music, photographs, video and other copyrighted works.
But the July 24 ruling should not be interpreted as a final answer. The court was deciding whether ANI was entitled to interim relief, not resolving every question surrounding copyright and AI training.
That leaves considerable room for the legal position to evolve.
Licensing Debate Is Far From Over
The immediate result is favourable to OpenAI because ANI was unable to secure the interim restrictions it sought. Yet the broader commercial dispute remains unresolved.
The most important question now is whether India’s courts eventually draw clear legal distinctions between training an AI model, storing copyrighted material, retrieving that material and reproducing it in an AI-generated response.
Those activities may carry different copyright implications, and the ANI case could help define where the boundaries lie.
For publishers, the case is a warning that simply producing original journalism may not guarantee control over how that information is used in the AI economy. For AI companies, meanwhile, the ruling offers some interim relief but does not remove the need for a sustainable approach to copyrighted data.
The next phase of India’s AI copyright debate is therefore unlikely to be about whether AI can learn from the internet at all. It will be about how the law balances technological innovation with the economic rights of the people and organisations that create the information AI systems learn from.
