The High Court of Delhi has dismissed an interim injunction application filed by ANI Media Pvt. Ltd. against Open AI, ruling that the use of copyrighted news content for training Large Language Models (LLMs) qualifies as fair dealing under the Copyright Act, 1957. The decision, delivered by Justice Amit Bansal on July 24, 2026, marks a major judicial precedent in India’s developing legal landscape concerning the intersection of artificial intelligence, data storage, and the boundaries of intellectual property protections.

In a comprehensive 135-page judgment, Justice Bansal concluded that ANI failed to demonstrate a prima facie case for copyright infringement, either in the initial training stage or in the generation of responses by ChatGPT. The Court observed that the technological processes underpinning modern AI—specifically the storage and tokenization of training data—do not violate the exclusive rights granted to copyright owners, provided those actions fall within the ambit of permissible use as defined by Section 52 of the Copyright Act.

The litigation centered on two specific claims: that OpenAI unlawfully stored ANI’s news articles during the model’s pre-training phase, and that ChatGPT later reproduced these works in its outputs, violating the news agency’s proprietary rights. Addressing the training claim, the Court found that the process of ingesting vast amounts of web-based data is an essential component of advancing machine learning and should not be severed from its broader utility in AI development.

“The process of training LLMs underlying ChatGPT can be considered as a form of research for generating new knowledge and advancing artificial intelligence systems,” Justice Bansal noted in his findings. The Court further clarified that the "purpose" and "fairness" tests were satisfied, as OpenAI's storage of data for training did not involve the unauthorized reproduction of expressive content for public display, but rather the internal analytical processing of linguistic patterns.

Regarding the output claim, the Court rejected the contention that ChatGPT memorizes and "regurgitates" ANI’s content in a way that constitutes infringement. Justice Bansal highlighted the technical nature of LLMs, explaining that the models predict sequences based on statistical distributions rather than maintaining a repository of raw source texts. Crucially, the judge pointed out that the specific articles cited by ANI in the lawsuit were published after the model's training cut-off dates, effectively debunking the argument that the responses were direct extractions of training data.

“In my prima facie view, Open AI’s acts of storage of the literary works of ANI for the training of its LLMs would fall under Section 52(1)(a) of the Copyright Act and therefore, does not amount to infringement under Section 51 of the Copyright Act,” the Court ruled.

The judgment extensively interpreted the scope of Section 52(1)(a)(i) , which allows for "private or personal use, including research." Rejecting the plaintiff’s narrow interpretation, Justice Bansal emphasized that the statutory protections for research are not limited to human-centric or non-commercial activity. Applying the " doctrine of updating construction ," the Court asserted that statutory language must be interpreted in light of current technological realities, affirming that machine learning processes constitute legitimate research under existing copyright frameworks.

Moreover, the Court addressed the balance of convenience, concluding that granting an injunction would stifle technological growth in India. Justice Bansal emphasized that the public interest—specifically the democratized access to information, educational tools, and advancements in scientific research—outweighed the private interests of a single news agency in this specific context. The Court underscored that ANI had failed to provide concrete evidence of market substitution or financial damage directly traceable to OpenAI’s training activities.

“Irreparable injury would be caused not only to Open AI but also to the public at large, if an interim injunction is granted in favour of ANI at this stage,” the judgment stated.

The legal analysis also touched upon the "fairness test," noting that the functions performed by generative AI are transformative in nature. Unlike a news aggregator that might compete directly for readership, ChatGPT acts as a tool for diverse intellectual tasks, such as summarizing, translating, and coding. Because the output generated by the LLM is fundamentally different from the original news reporting in style, character, and intent, the Court held it could not be considered a substitute for the plaintiff’s syndicated content.

Justice Bansal also dismissed concerns regarding the lack of explicit licensing agreements, noting that ANI possessed the technical ability to "opt-out" of OpenAI’s web crawlers but had chosen not to exercise that control over its digital footprints until the litigation commenced. The Court further remarked that because ANI had previously offered a license to OpenAI for a specified fee, the alleged damages were purely monetary and compensable, further weakening the grounds for an extraordinary equitable remedy like an interim injunction.

The dismissal of I.A. 45300/2024 serves as a pivotal moment for Indian AI developers and global tech firms, signaling that the judiciary is prepared to adopt a liberal, purposive approach to copyright law. By prioritizing scientific advancement and technological innovation, the Court has provided a degree of legal certainty for startups and research labs currently training domestic LLMs. While the main suit (CS(COMM) 1028/2024) remains pending, the Court has firmly established that the mere training of algorithms on publicly available, non-infringing digital data does not constitute a copyright violation under the present Indian legal framework. The parties are now expected to proceed to trial, where further evidence will be submitted to test the long-term implications of these findings.