On Friday, the court turned down the request of Asian News International (ANI) for an interim injunction which would essentially stop the unauthorised use of its content to train OpenAI’s Large Language Models (LLMs) for generating responses.
LLMs are deep-learning models trained on vast amounts of publicly available text data, like articles, websites, and books. They also generate human-like responses to queries, which may not always be accurate.
“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 Amit Bansal ruled, two years after the news agency took the parent company behind ChatGPT to court.
Although the main suit is still pending, the application for interim injunction has been dismissed at this stage.
One of the major reasons for rejection of the plea was that OpenAI’s act of storing ANI’s literary works for training its LLMs fell under the “fair dealing doctrine” which has been mentioned under Section 52 of the Act.
Section 52 provides a comprehensive list of specific acts that do not constitute copyright infringement such as using copyrighted works for personal research, reporting, private use, criticism or review of any other work, among others.
ANI, the court said, has failed to make out a prima facie case for the grant of an interim injunction.
Saying that the grant of an interim injunction would hinder AI’s growth and more particularly of any LLMs developed in India, the court said, allowing it would have an “adverse impact on public interest” and on millions of users of ChatGPT in India, many of whom are not paid subscribers.
“Public interest is an important aspect…that the courts have to consider while granting interim injunction, more particularly in intellectual property cases.”
Explaining that the development and success of LLMs depend on data availability, the court said that it would be economically unviable to develop an LLM, if training it required licenses from multiple sources.
Importantly, the court rejected the plea for reasons such as ANI’s admitted position that it has the ability to block any website vis-a-vis any third party, including OpenAI.
“The opting out option is available to ANI for blocking the third-party web crawlers from copying their data as well as from scraping their website for the search function/RAG. Despite having an option of opt-out, evidently ANI has not exercised the same,” the court noted in its 125-page order.
The news agency had contended that there were various paid subscribers, whose websites carry its copyrighted data, and Open AI can still access and copy those data from those websites.
“Even if that be so, it would be open for the subscribers also to block their websites to Open AI. Clearly, if they have chosen not to do so, it could be possibly for their own commercial reasons,” the court noted in its order.
Also Read: AI-generated art covered by India’s copyright law? Delhi HC gives copyright office 8 weeks to decide
Understanding copyright
A “copyright” is a right given by the law to creators of literary, dramatic, musical, and artistic works and producers of cinematographic films and sound recordings. Simply put, it is a bundle of rights, including the right of reproduction, communication to the public, adaptation, and translation of the work. The 1957 Act seeks to protect such creative works, which are deemed as the creator’s intellectual property (IP).
Usually, when a copyright is infringed, or used in an unauthorised manner, courts can step in to provide relief by restraining the use of another’s copyright through ad-interim (temporary) or permanent injunctions. They can also order damages in favour of those affected by the infringement.
“ANI has failed to make out a case for prima facie grant of interim injunction. The balance of convenience is also against the grant of such injunction. Allowing the plea for interim injunction would cause irreparable injury to not just OpenAI but also to the public at large,” the court said.
Significantly, the court said that the petitioner failed to satisfy it about the regurgitation or memorisation of ANI’s original literary responses which were being generated by the chatbot, ChatGPT.
Underlining that the OpenAI outputs were not substantially similar to ANI’s works, the court said that the inputs generated by the chatbot using the Retrieval-Augmented Generation (RAG) technique did not amount to infringement of copyright. The RAG technique usually enhances the LLMs by pulling facts from one’s private data to ground responses.
“The case is certainly the first-of-its kind in the country, and will shape future discourse on the topic in the future,” advocate Adarsh Ramanujan, who appeared as an amicus in the matter, told ThePrint.
Importantly, the present order will have no precedential value, as it is not a final judgment. Despite this, the case is the first-of-its-kind and engages with cutting-edge issues nonetheless, Ramanujan said.
The story so far
In November 2024, the Delhi HC had issued summons to Open AI on the plea by ANI which alleged unauthorised use of its content to train its LLM to generate human-like responses to queries.
ANI also sought an injunction to prevent such unauthorised use of its content by the chatbot. An injunction is “an official order given by a court, usually to stop someone from doing something.
OpenAI said it blocked the news agency’s website to prevent the chatbot from using its content, a submission that the court went on to record in its November 19 order.
ANI had moved the Delhi High Court in November, 2024, alleging unjust enrichment, unfair competition, and other wrongs by OpenAI.
Submitting that it has built a large archive made up of text, exclusive statements, articles, programmes, interviews, videos, and images over the last five decades, ANI said that it is the exclusive owner of the copyright over this content under the Copyright Act.
OpenAI, it contended, developed the Al language model platform Chat Generative Pre-Trained Transformer (ChatGPT), which is an interactive, conversational text-based Al platform based on a large-language model that enables it to respond to queries by users.
Pointing out that these large-language models are developed, trained, and attuned by “feeding it a database of information,” including its works, ANI said that ChatGPT is a commercial entity.
Such content remains in ChatGPT’s memory indefinitely, ANI said adding that the sole purpose of such storage and use is to bolster the chatbot’s capability and functioning.
Using and storing ANI’s works, without license or permission, to provide services to their users and even paid subscribers, violates exclusive right, it said. Moreover, OpenAI had “in many instances, falsely attributed statements” to ANI that misled the general public.
“All these acts are actionable in law, including for infringement of copyright, other intellectual property rights, disparagement, unfair competition, and unjust enrichment,” the petition submitted.
ANI sought to restrain Open AI from accessing, storing, reproducing, publishing, or in any other manner, using its content, on all versions of ChatGPT. In addition, it sought damages suffered on account of copyright infringement, unjust enrichment, unfair competition, and other wrongs.
Open AI was indulging in unfair competition by using its content to publish and timely report accurate information on current affairs to its own subscribers, which could not only drive away ANI’s readers but also affect its entire business model and operations, the petitioner contended.
(Edited by Tony Rai)
Also Read: Why Delhi HC’s pecuniary jurisdiction proposal has triggered strong pushback from IP lawyers