Introduction:
In ANI Media Pvt. Ltd. v. OpenAI Inc. & Another, the Delhi High Court delivered a landmark interim ruling in India’s first copyright dispute involving a generative artificial intelligence platform. Justice Amit Bansal dismissed the interim injunction application filed by ANI Media Pvt. Ltd. against OpenAI Inc., the developer of ChatGPT, while holding that, on a prima facie assessment, OpenAI’s use of ANI’s copyrighted material for training its AI model does not amount to copyright infringement under the Copyright Act, 1957. The Court also held that the responses generated by ChatGPT were not substantially similar to ANI’s original works and therefore did not prima facie infringe copyright.
Summons in the suit had earlier been issued in November 2024. ANI instituted the suit alleging that OpenAI was commercially exploiting its original news reports by using them to train ChatGPT and that the chatbot was capable of reproducing ANI’s copyrighted content verbatim in response to user prompts. The news agency further alleged that ChatGPT occasionally attributed false reports to ANI, resulting in so-called “hallucinations” that threatened its journalistic credibility and contributed to the spread of misinformation.
Apart from the copyright dispute, the case also involved an important jurisdictional challenge. OpenAI argued that Indian courts lacked territorial jurisdiction because its AI models were trained outside India and its servers were located in the United States. However, the Delhi High Court ruled, at the prima facie stage, that it possessed territorial jurisdiction to entertain the suit. While upholding jurisdiction, the Court ultimately declined to grant the interim injunction sought by ANI.
Arguments of the Parties:
ANI contended that it owns copyright over its original news reports, articles and journalistic content. It alleged that OpenAI had copied, stored and used its copyrighted material without permission for training ChatGPT, thereby commercially benefiting from ANI’s intellectual property.
According to ANI, ChatGPT was capable of reproducing portions of its news reports almost verbatim when responding to user queries. Such reproduction, it argued, amounted to direct infringement of its exclusive copyright under the Copyright Act, 1957. ANI further submitted that OpenAI’s commercial AI model was built upon unauthorised use of copyrighted content belonging to various news organisations.
The plaintiff also highlighted instances where ChatGPT allegedly attributed fabricated news reports or statements to ANI. It argued that these AI-generated “hallucinations” posed a serious threat to the reputation of a professional news agency and could mislead the public by spreading inaccurate information under ANI’s name.
Seeking interim relief, ANI requested the Court to restrain OpenAI from storing, processing or using its copyrighted material during the pendency of the suit, contending that continued use would cause irreparable injury to its proprietary and commercial interests.
OpenAI opposed the application on both jurisdictional and substantive grounds. It initially challenged the maintainability of the suit in India by arguing that its AI models are trained outside the country and that the relevant technological infrastructure is located in the United States.
On merits, OpenAI submitted that AI training is fundamentally a transformative process rather than a process of reproducing copyrighted expression. It argued that machine learning analyses enormous datasets to understand linguistic relationships and statistical patterns rather than storing copyrighted works for republication.
The company further contended that copyright protects only the original expression of ideas and not the underlying facts contained in news reports. Since factual information cannot be monopolised through copyright law, the use of factual material for AI training does not amount to infringement.
OpenAI also submitted that publishers who do not wish to have their publicly accessible content crawled for AI purposes may opt for technical blocklisting mechanisms.
The Court also received valuable assistance from the amici curiae. Professor Dr. Arul George Scaria of the National Law School of India University argued that the High Court possessed territorial jurisdiction to decide the dispute. He further distinguished between expressive and non-expressive use of copyrighted material, submitting that OpenAI’s use of ANI’s content was, in most cases, non-expressive and therefore did not amount to copyright infringement. Advocate Adarsh Ramanujan also assisted the Court as amicus curiae.
Court’s Judgment:
Justice Amit Bansal first held that the Delhi High Court possessed territorial jurisdiction to entertain the suit, rejecting OpenAI’s preliminary objection at the prima facie stage. Having found the suit maintainable, the Court proceeded to consider whether ANI had established the requirements necessary for grant of an interim injunction.
The Court observed that, on a prima facie examination, OpenAI’s act of storing ANI’s original works during the AI training process falls within the protection available under Section 52(1)(a) of the Copyright Act, 1957. This provision recognises the doctrine of fair dealing and excludes certain categories of use from the scope of copyright infringement.
Accordingly, the Court held that the storage of ANI’s copyrighted material for AI training did not prima facie constitute infringement under Section 51 of the Copyright Act.
The Court then examined ANI’s allegation that ChatGPT reproduced its copyrighted works. Justice Bansal observed that the outputs generated by ChatGPT were not substantially similar to ANI’s original reports. Since copyright infringement requires reproduction of a substantial part of the protected expression, the Court found that ANI had failed to establish a prima facie case.
The Court further observed that ANI had not demonstrated any instance of “memorisation” or “regurgitation” of its copyrighted works by ChatGPT. In other words, the material placed before the Court did not establish that the AI system was reproducing ANI’s original content in a manner amounting to copyright infringement.
Justice Bansal therefore concluded that the plaintiff had failed to satisfy the first requirement for grant of interim relief, namely the existence of a strong prima facie case. The Court also considered the balance of convenience and the likelihood of irreparable injury.
According to the Court, granting an interim injunction at this stage would cause serious prejudice not only to OpenAI but also to the public at large by potentially restricting the functioning and development of an important artificial intelligence platform before the legal issues were finally adjudicated. On the other hand, the Court found that the plaintiff had not established immediate irreparable harm warranting such an extraordinary interim order.
Accordingly, the interim injunction application was dismissed while leaving all questions to be finally decided during trial after examination of complete evidence and detailed legal arguments.
The ruling marks a significant development in Indian copyright jurisprudence concerning artificial intelligence. It recognises, at least at the interim stage, that AI model training and AI-generated outputs require careful legal analysis distinct from conventional copyright disputes. The judgment also reflects the Court’s attempt to balance the interests of copyright owners with technological innovation and public interest.
Although the Court declined interim relief, the copyright suit itself remains pending. The final judgment is expected to address several complex issues, including the legality of AI training on copyrighted material, the scope of fair dealing under Indian copyright law, the distinction between expressive and non-expressive use, the legal consequences of AI-generated hallucinations and the evolving relationship between artificial intelligence and intellectual property rights in India.