# Indian Court Rules OpenAI Did Not Infringe ANI Copyright in Training > Source: > Published: 2026-07-24 11:29:11+00:00 **July 24, 2026**, (Inside AI) — A New Delhi court ruled today that OpenAI did not infringe copyright when it used news content from Asian News International (ANI) to train its ChatGPT service, delivering a landmark verdict in a case closely watched by AI developers and publishers worldwide. The decision marks a significant legal validation of using publicly available data for training large language models under Indian law. The court found that OpenAI's ingestion of ANI's articles constituted fair dealing, not infringement, setting a precedent that could ripple across other jurisdictions grappling with similar disputes. Justice Pratibha M. Singh of the Delhi High Court stated: **"The use of copyrighted works for the purpose of training an AI model, when such works are lawfully accessible, does not constitute infringement under Section 52 of the Copyright Act, 1957, as it falls within the ambit of fair dealing for research and technological development."** Justice Pratibha M. Singh, Delhi High Court The lawsuit, filed by ANI in 2024, alleged that OpenAI scraped its proprietary news content without license to train ChatGPT, thereby violating its copyright and undermining its business model. ANI argued that ChatGPT could reproduce near-verbatim excerpts of its articles, effectively competing with the news agency's own services. OpenAI countered that its models learn patterns from data rather than storing or copying works, and that training on publicly available information is a transformative fair use. The company also noted that it had implemented technical measures to prevent verbatim regurgitation of training data. The ruling aligns with similar outcomes in the United States, where courts have generally favored AI developers in fair use cases, though several lawsuits remain ongoing. However, it diverges from the European Union's more restrictive approach under the AI Act, which requires transparency about training data and offers opt-out mechanisms for rights holders. ## How the Verdict Reshapes AI Training Boundaries The Delhi court's reasoning leaned heavily on the concept of non-expressive use, a doctrine that permits copying when the purpose is not to convey the original expression to users. This echoes the U.S. *Authors Guild v. Google* case, where digitizing books for search was deemed fair use. Legal experts note that the Indian Copyright Act's fair dealing provision is narrower than the U.S. fair use doctrine, yet the court interpreted it broadly to accommodate technological innovation. The judgment emphasized that training AI on copyrighted works serves a public interest by advancing scientific progress. However, the court cautioned that if an AI model is used to generate outputs that directly compete with the original work or replicate substantial portions, that could constitute infringement. This leaves the door open for future claims based on specific outputs rather than the training process itself. ANI's legal team argued that OpenAI's scraping violated the terms of service of its website, which prohibit automated access. The court dismissed this, stating that robots.txt protocols are not legally binding under Indian law and that OpenAI had not circumvented any technical protection measures. ## Global Ripple Effects and Unanswered Questions The verdict arrives amid a flurry of copyright battles against AI firms. In 2025, the U.S. Copyright Office issued guidance affirming that AI training on copyrighted works may be fair use, but it stopped short of a blanket exemption. Meanwhile, Japan and Singapore have enacted explicit text-and-data-mining exceptions to copyright law. OpenAI welcomed the ruling, with a spokesperson saying it "validates our approach to building AI that benefits everyone while respecting creators' rights." ANI has indicated it will appeal, setting up a potential Supreme Court showdown. The decision does not address the separate issue of whether AI-generated outputs can be copyrighted, a matter currently under deliberation by Indian policymakers. It also leaves unresolved the question of compensation for publishers, a topic being negotiated in licensing deals between AI firms and major news organizations like the Associated Press and Axel Springer. For now, AI developers in India can rely on this precedent to continue training on publicly accessible web data, though they must remain vigilant about output filtering to avoid direct competition with original works. The ruling may also influence pending cases in the UK and Canada, where courts are considering similar arguments.