{"slug": "delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training", "title": "Delhi High Court Rules No Copyright Infringement in OpenAI AI Training", "summary": "The Delhi High Court ruled that OpenAI did not commit copyright infringement when using publicly available texts to train its AI models, dismissing a lawsuit by the Federation of Indian Publishers. Justice Pratibha M. Singh found the training process falls under fair dealing for research and analysis, as the models learn statistical patterns without reproducing protected expression. The decision aligns India with U.S. fair use precedents but diverges from EU rules, potentially accelerating AI development in India while leaving output-specific infringement claims open.", "body_md": "**July 24, 2026**, (Inside AI) — The Delhi High Court has ruled that OpenAI did not commit copyright infringement when using publicly available texts to train its artificial intelligence models. The decision, delivered by Justice Pratibha M. Singh, dismisses a lawsuit filed by the Federation of Indian Publishers, which argued that the company’s scraping of books and articles without licenses violated Indian copyright law.\n\nThe court found that the training process falls under the fair dealing exception for research and analysis in Section 52 of the Copyright Act, 1957. Justice Singh noted that the AI training constituted a non-expressive, transformative use of the works, as the models learn statistical patterns rather than reproducing protected expression.\n\n**\"The use of copyrighted works for training AI models, where the output does not replicate the original expression, is akin to reading a book to learn facts and ideas, which copyright does not monopolize,\"** Justice Pratibha M. Singh, Delhi High Court.\n\nThe ruling aligns India with jurisdictions like the United States, where courts have leaned toward fair use in similar cases, such as *Authors Guild v. Google*. However, it diverges from the European Union’s stricter text and data mining exceptions, which require opt-out mechanisms for rights holders. This decision is likely to accelerate AI development in India, giving companies a legal safe harbor to train on local language data without licensing hurdles.\n\nYet, the judgment sidesteps the contentious issue of whether AI outputs can sometimes infringe. The court clarified that if a model generates content substantially similar to a copyrighted work, that specific output could still be actionable. This leaves the door open for future litigation focused on model outputs rather than training inputs.\n\nThe Federation of Indian Publishers expressed disappointment, arguing that the decision undervalues creative labor. They plan to appeal, potentially taking the case to the Supreme Court. Globally, this ruling adds to a patchwork of legal standards, with similar cases pending in the UK and China. For now, India positions itself as a favorable jurisdiction for AI training, but the debate over creator compensation remains far from settled.\n\n## Training Data and Fair Use: A Global Patchwork\n\nThe Delhi High Court’s reasoning draws heavily from the idea-expression dichotomy, a bedrock of copyright law. By treating AI training as the extraction of unprotected facts and ideas, the court echoed the U.S. Supreme Court’s 1991 decision in *Feist Publications v. Rural Telephone Service*, which held that facts are not copyrightable. This approach is detailed in a [recent paper on fair use and generative AI](https://arxiv.org/abs/2304.09388), which argues that training on copyrighted data is transformative when the model does not store or reproduce the works.\n\nHowever, critics point out that large language models can memorize and regurgitate verbatim text, as shown in a [study on memorization in LLMs](https://arxiv.org/abs/2212.03860). The court acknowledged this risk but placed the burden on plaintiffs to prove infringement on a case-by-case basis for outputs. This shifts the legal battle from training to deployment, where monitoring every generated sentence becomes impractical.\n\n## India’s AI Ambitions and the Publisher Backlash\n\nThe ruling comes as India pushes to become an AI powerhouse, with initiatives like the IndiaAI Mission. By providing legal clarity, the court may attract more AI investment, but it also risks alienating the publishing industry. Publishers argue that without licensing, their content is devalued, threatening the production of quality journalism and literature. The government may need to step in with a statutory licensing scheme, similar to what is being debated in the UK, to balance innovation and creator rights.\n\nFor now, the decision is a significant win for OpenAI and other AI companies operating in India. But as AI-generated content floods the market, the line between inspiration and infringement will only blur further, ensuring that this is just the opening chapter in a long legal saga.", "url": "https://wpnews.pro/news/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training", "canonical_source": "https://insideai.news/news/ai-policy-and-regulation/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training/5217/", "published_at": "2026-07-24 11:34:36+00:00", "updated_at": "2026-07-24 11:38:47.185154+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-research", "ai-ethics"], "entities": ["Delhi High Court", "OpenAI", "Federation of Indian Publishers", "Justice Pratibha M. Singh", "India", "United States", "European Union"], "alternates": {"html": "https://wpnews.pro/news/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training", "markdown": "https://wpnews.pro/news/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training.md", "text": "https://wpnews.pro/news/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training.txt", "jsonld": "https://wpnews.pro/news/delhi-high-court-rules-no-copyright-infringement-in-openai-ai-training.jsonld"}}