# Delhi High Court Rules AI Training Is 'Fair Dealing,' Rejects Indian News Agency's Injunction Against OpenAI > Source: > Published: 2026-07-28 11:22:13.579383+00:00 # Delhi High Court Rules AI Training Is 'Fair Dealing,' Rejects Indian News Agency's Injunction Against OpenAI - Justice Amit Bansal ruled that OpenAI's storage of ANI's works for training LLMs falls under Section 52(1)(a) of the Indian Copyright Act and does not amount to infringement [[1]](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/) - ANI failed to demonstrate that ChatGPT memorized or reproduced its articles — key evidence cited articles published after model training cutoff dates [[2]](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/) - The court warned that granting an injunction would cause 'irreparable injury' to OpenAI and the public, noting India's 100 million ChatGPT users [[3]](https://www.livelaw.in/high-court/delhi-high-court/delhi-high-court-rejects-ani-copyright-suit-open-ai-chatgpt-542736) - This is the third major fair-use/fair-dealing ruling favoring AI labs in 13 months, following two US federal court decisions in June 2025 [[4]](https://thenewpublishingstandard.com/2026/07/27/delhi-high-court-openai-ani-fair-dealing-ruling/) - The ruling is a prima facie finding only; the full copyright trial between ANI and OpenAI remains pending [[1]](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/) India's Delhi High Court on July 24 rejected a bid by Asian News International, the country's largest news wire service, to bar OpenAI from using its copyrighted articles to train ChatGPT. Justice Amit Bansal ruled in a 135-page order — delivered after 32 hearings — that OpenAI's ingestion of ANI's content for model training constitutes "fair dealing" under Section 52(1)(a) of India's Copyright Act, the first time any court globally has classified AI training as permissible private use [1]. "I am of the prima facie view that Open AI's act of storing ANI's original literary works for training LLMs underlying ChatGPT falls under Section 52(1)(a) of the Copyright Act and therefore, does not amount to infringement," Justice Bansal wrote [3]. The judge found that ChatGPT's outputs were not "substantially similar" to ANI's original reporting and that the news agency had failed to prove any memorization or regurgitation of its work . [[3]](https://www.livelaw.in/high-court/delhi-high-court/delhi-high-court-rejects-ani-copyright-suit-open-ai-chatgpt-542736)The decision lands as the third major judicial win for AI companies on copyright training in just over a year, following two US federal court rulings in June 2025 — Bartz v. Anthropic and Kadrey v. Meta — that similarly favored AI developers [4]. Legal scholars say the convergence across India's exhaustive statutory fair-dealing framework and America's flexible fair-use doctrine suggests a forming global judicial consensus on the issue . [[4]](https://thenewpublishingstandard.com/2026/07/27/delhi-high-court-openai-ani-fair-dealing-ruling/)## The Ruling Justice Bansal applied a three-part test to ANI's injunction request and found OpenAI prevailed on every count. First, the court determined that ANI had not established a prima facie case of copyright infringement because ChatGPT's outputs bore no substantial similarity to the agency's original articles [2]. Second, the balance of convenience favored OpenAI. Third, granting relief would harm the broader public interest . [[3]](https://www.livelaw.in/high-court/delhi-high-court/delhi-high-court-rejects-ani-copyright-suit-open-ai-chatgpt-542736)A critical flaw in ANI's evidentiary case proved decisive: the news agency cited articles published in August and September 2024 as proof of infringement, but GPT-4 was trained on data through April 2022 and GPT-4o through April 2024 — meaning the articles could not have been in the training data [2]. The court attributed any similarities in ChatGPT's responses to Retrieval Augmented Generation (RAG), a technique that fetches real-time web information rather than drawing on memorized training data . [[2]](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/)The judge also rejected ANI's argument that OpenAI's commercial status should automatically disqualify it from fair-dealing protections, closing off a strategy publishers had pursued in multiple jurisdictions [4]. ## Why It Matters The ruling carries outsized significance because India's Copyright Act uses an exhaustive statutory list of permitted uses — unlike the flexible, four-factor fair-use balancing test in US law. By finding AI training fits within Section 52(1)(a)'s language of 'private or personal use, including research,' the Delhi High Court extended fair-dealing protections to a use case the statute's drafters never contemplated [1]. The decision also established that Indian courts have jurisdiction over AI training conducted on US-based servers, preventing companies from using server location to sidestep Indian copyright law [4]. At the same time, the court weighed public interest heavily, warning that requiring licenses before training would make LLM development 'economically unviable' and that an interim injunction would 'hinder AI growth and more particularly of any LLMs developed in India' . [[5]](https://theprint.in/judiciary/no-interim-ban-on-openai-delhi-high-court-invokes-fair-dealing-doctrine-to-reject-anis-petition/2996430/)For publishers and news organizations worldwide, the ruling shifts the burden of proof. Rather than demonstrating that their content was ingested — which is generally undisputed — rights holders must now show actual memorization or substantial reproduction in model outputs, a significantly higher evidentiary bar [4]. ## Global Context The Delhi decision follows two US federal court rulings in the Northern District of California in June 2025: Judge William Alsup dismissed key claims in Bartz v. Anthropic on June 23, and Judge Vince Chhabria did likewise in Kadrey v. Meta two days later [4]. All three rulings reached similar conclusions despite operating under fundamentally different legal frameworks. The New York Times lawsuit against OpenAI, filed in December 2023, remains the highest-profile unresolved case in the space. Thomson Reuters' suit against Ross Intelligence, which settled, also addressed AI training on copyrighted text. The Delhi ruling adds momentum to AI companies' legal position ahead of those remaining proceedings [4]. ANI had originally filed its suit against OpenAI in November 2024, alleging unjust enrichment, unfair competition, and multiple intellectual property violations [6]. The news agency argued that ChatGPT's outputs amounted to 'exploitation for commercial gain' and that AI-generated 'hallucinations' attributed to ANI damaged the agency's credibility . [[3]](https://www.livelaw.in/high-court/delhi-high-court/delhi-high-court-rejects-ani-copyright-suit-open-ai-chatgpt-542736)## What's Next The ruling is explicitly a prima facie finding and will not control the final outcome of the case. The full copyright trial between ANI and OpenAI remains pending before the Delhi High Court, where discovery, expert evidence, and deeper examination of RAG-based outputs await [1]. The court left open the question of whether RAG-generated outputs — which pull from live web sources rather than training data — constitute 'communication to the public' under Indian copyright law, a distinct legal issue that will be addressed at trial [2]. For the broader AI industry, the ruling provides significant — if preliminary — legal cover for training on copyrighted material across jurisdictions. Combined with the two US rulings from 2025, AI companies now have favorable interim precedent on three continents, though no final judgment has yet been rendered in any major AI copyright case [4]. ## Companies mentioned ## Further sources [[1] The Decoder — Delhi High Court hands OpenAI a win by rejecting major Indian new… ↗](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/) [[2] The Decoder — detailed analysis of evidence and RAG findings in ANI v. OpenAI r… ↗](https://the-decoder.com/delhi-high-court-hands-openai-a-win-by-rejecting-major-indian-news-agencys-copyright-injunction/) [[3] LiveLaw — Delhi High Court Rejects ANI's Interim Plea Against ChatGPT, Says It … ↗](https://www.livelaw.in/high-court/delhi-high-court/delhi-high-court-rejects-ani-copyright-suit-open-ai-chatgpt-542736) [[4] The New Publishing Standard — Delhi High Court Rules for OpenAI: The Third Majo… ↗](https://thenewpublishingstandard.com/2026/07/27/delhi-high-court-openai-ani-fair-dealing-ruling/) [[5] ThePrint — No interim ban on OpenAI: Delhi High Court invokes fair dealing doct… ↗](https://theprint.in/judiciary/no-interim-ban-on-openai-delhi-high-court-invokes-fair-dealing-doctrine-to-reject-anis-petition/2996430/) [[6] EU Intellectual Property Helpdesk — India's news website ANI sues OpenAI allegi… ↗](https://intellectual-property-helpdesk.ec.europa.eu/news-events/news/indias-news-website-ani-sues-openai-alleging-copyright-infringement-2024-12-04_en) The stories that matter, in one email. Free — unsubscribe anytime.