Delhi HC dismisses allegations of copyright infringement against ChatGPT

New Delhi: In an interim relief to OpenAI, the Delhi High Court on Friday dismissed allegations of copyright infringement against the ChatGPT parent, rejecting the lawsuit filed by news agency Asian News International (ANI), publishers' bodies and the Indian music industry.

According to the judgment reviewed by Mint, the Delhi High Court's findings are only prima facie and confined to deciding ANI's application for interim relief. The copyright suit will continue, and the observations do not amount to a final determination on whether OpenAI infringed ANI's copyright.

The court prima facie held that while OpenAI temporarily stores copyrighted literary works during the training of its AI models, such storage is protected as fair dealing under the Copyright Act and does not amount to copyright infringement at the interim stage.

It found no copyright infringement in the San Francisco-based company's use of copyrighted material to train its large language models (LLMs), holding that such use falls within the exceptions under Section 52 of the Copyright Act.

It ruled that ChatGPT's outputs were not substantially similar to ANI's copyrighted works and that the news agency had failed to establish that the chatbot memorised or reproduced its original content.

The court said granting an interim injunction against OpenAI would cause irreparable harm not only to the company but also to the public, which benefits from AI services.

The court clarified that Indian courts have jurisdiction to hear the dispute despite OpenAI's servers being located in the US, holding that the alleged infringement had a sufficient connection with India because ChatGPT is accessible in the country and the alleged harm was suffered by ANI in India.

However, the ruling found that the examples cited by ANI did not establish that ChatGPT had substantially reproduced ANI's copyrighted news reports. Further, ANI had failed to demonstrate any actual market harm or loss of subscriptions arising from OpenAI's use of its content.

Rejecting ANI's plea for an interim injunction, the court held that the balance of convenience favoured OpenAI and observed that generative AI technologies serve a significant public interest by supporting research, education, accessibility and the dissemination of knowledge.

The court also rejected the submissions of the Federation of Indian Publishers (FIP), the Digital News Publishers Association (DNPA) and the Indian music industry (IMI) at the interim stage, declining to accept their contention that storing and using copyrighted works for AI training necessarily amounts to copyright infringement.

The members of FIP and the DNPA include HT Digital/HT Media (publisher of Mint and Hindustan Times), The Hindu, India Today, Network18, NDTV, ABP Group, Dainik Jagran, and Times Internet. T-Series, Saregama India, and Sony Music India were the companies in the music industry that backed the plea.

According to a lawyer familiar with the matter, the case was a "test of the waters" as it was India's first major AI copyright dispute, particularly for publishers and music industry bodies that intervened in the proceedings.

The person, who requested anonymity, said the interim order is likely to prompt publishers and music companies to file separate copyright infringement suits against OpenAI in the coming months, as they believe they have stronger claims based on the use of their copyrighted content

Ankit Sahni, counsel for the Federation of Indian Publishers (FIP), the first intervenor in the case, said the judgment departs from the policy direction being considered by the government. He noted that a DPIIT-appointed committee had recommended a compulsory licensing framework under which AI companies could use copyrighted works to train their models by paying licence fees to copyright owners.

Sahni explained that while the Delhi High Court held that training AI models on copyrighted material is protected under the Copyright Act's fair dealing provisions and does not amount to copyright infringement, which is contrary to the government's approach.

Ronil Goger, managing partner at Blaze Legal, said ANI can appeal the interim order before a division bench of the Delhi High Court while the main suit continues, where the court will examine technical evidence on AI training and copyright infringement. "Given the novel and far-reaching questions involved, the dispute is likely to ultimately be considered by the Supreme Court," he said.

Replying to Mint’s query, news agency ANI said, “After reviewing the judgement, we will consider our legal options, including any potential appeals or future frameworks, in due course and in consultation with our legal advisors. We have no further comments to offer at this time.”

Queries emailed to OpenAI remained unanswered till press time.

What happened earlier

The dispute began in November 2024 when ANI sued OpenAI, alleging it used the agency's copyrighted news reports without permission to train ChatGPT. The case later expanded with interventions from publishers and music industry bodies, which argued that using copyrighted works for AI training without a licence infringes intellectual property rights.

OpenAI has denied the allegations, saying its models are trained on publicly available data and do not reproduce copyrighted works.