Delhi High Court has ChatGPT’s parent company, OpenAI, its first win in a copyright row, ruling that the data used to train its large language model (LLM) is publicly available and considered “fair dealing” under copyright law.
The court announced its judgment on 24 July 2026 in the case of . Indian media company Asian News International (ANI) filed this application, seeking an interim injunction against OpenAI. The court dismissed the application and its judgment did not apply to the main suit, which is pending disposal.
The question before the court was whether the use of publicly available data by AI innovators for training LLMs and its subsequent production as outputs to user prompts, all without the permission of the data owners, amounted to copyright infringement, or fair dealing.
ANI argued that ChatGPT’s outputs used both original and creative elements of ANI’s text, thus violating the media company’s copyright.
It also argued that some responses were near and/or exact copies of ANI’s data. The generated responses had no application of mind, and at best could be considered an adaptation of ANI’s work as the responses were mere rearrangements of the original, the media company said. ANI also said OpenAI had admitted verbatim reproduction.
ANI submitted that OpenAI had recognised the proprietary rights of other news organisations and entered into licensing agreements with them.
The media company also argued that public availability of its data did not nullify its copyright or grant a universal licence for use. ANI also submitted that OpenAI’s actions did not fall under “private or personal use, including research” protected under the law, as the company used the data for commercial purposes and profit.
ANI relied on the cases of , Advance Local Media LLC et al v Coheree Inc (2025), both in the US, and GEMA v OpenAI (2025) in Germany to support its arguments.
OpenAI argued that ANI’s work was available in the public domain and could be used without obtaining a licence. The cut-off date for training its LLM was several months prior to the AI model’s launch. The works at the centre of ANI’s copyright infringement allegations were published after the cut-off date and were not part of the training data for OpenAI’s LLM.
It was also argued that ChatGPT did not reproduce extracts, rather it used the information it had learnt from the data to respond to prompts.
OpenAI argued that the law protected the manner of expression of an idea or fact, not the idea or fact itself. For news materials, as is in this case, the underlying fact covered in the news and being used by the model, is not protected by copyright law and ANI cannot claim a monopoly over the facts.
OpenAI relied on the US case, , to submit that copyright in news was limited to only expression and in the case of news, the standard to establish similarity is higher.
OpenAI also relied on , to submit that Indian copyright laws adopt the “skill and judgement test” to determine whether a work is entitled to copyright protection. Under this test, a work must be more than just a copy, may not be creative but has to have the skill and judgement of its creator.
The defendants relied on and to submit that similarities alongside broad dissimilarities negated the intention of copying the original work, and where the coincidences were incidental, no copyright breach existed.
The court said in its analysis that ANI would have ownership and copyright over its original literary works, even if they were freely and publicly available.
The court, relying on and several other cases, said that “in the context of news, copyright would subsist only in the form and manner of expression of news and not on the underlying facts”. Prima facie, copyright infringement in ChatGPT responses was not established, it said.
The court relied on to reiterate that commercial use of a copyrighted work would not make it unfair or exclude it from the ambit of fair dealing.
The court added that “OpenAI’s storage and use of training data falls within the ambit of ‘private use’, including ‘research’”, which was protected under the law and did not amount to infringement.
