Back

AI Training, Copyright and Fair Dealing: ANI Media Private Limited v. OpenAI Opco LLC ((CS(COMM) 1028/2024)

Blog Post Thumbnail

In ANI Media Private Limited (“ANI”) vs. OpenAI OpCo LLC (“OpenAI”), the Hon’ble Delhi High Court (“Court”) considered whether OpenAI’s use and storage of ANI’s copyrighted news content for training the large language models (“LLMs”) underlying ChatGPT and the generation of outputs based on such training, amounted to copyright infringement under Section 51 of the Copyright Act, 1957 (“Act”).

The Court examined whether ChatGPT memorises and reproduces ANI’s works. It noted that copyright protects the expression of ideas and facts, rather than facts or information themselves. This distinction is particularly relevant to news content, which is substantially factual. The Court further observed that infringement requires substantial or material reproduction of protected expression. On the material placed before it, ANI had not demonstrated substantial similarity or verbatim reproduction in the cited examples. The Court also noted that memorisation and regurgitation may occur in LLMs in limited circumstances, but ANI had not established that such memorisation of its works had occurred in the responses generated by ChatGPT.

The Court accepted that OpenAI electronically stores ANI’s works, at least temporarily, during the LLM-training process. The Court held that Section 52 of the Act must be interpreted broadly as part of the statutory balance between copyright protection and public interest in creativity, research and dissemination of knowledge. The Court formulated a two-stage test: (i) whether the use falls within a purpose specified in Section 52(1)(a) of the Act; and (ii) whether the dealing is fair. Applying this test, the Court held that both the purpose test and fairness test were satisfied. It found that commerciality, by itself, does not exclude the use from Section 52(1)(a)(i) and that LLM training could fall within the concept of “private or personal use, including research”. It also found no sufficient material demonstrating that OpenAI’s use had affected the market share or caused potential or actual damage to ANI

Accordingly, the Court held that storage of ANI’s works for LLM training fell within Section 52(1)(a) and did not amount to infringement. It also found no prima facie infringement in the alleged outputs or retrieval-augmented generated responses. Consequently, ANI failed to establish a prima facie case for interim injunction, and the application was dismissed. The Court clarified that its observations were confined to the interim stage and would not determine the final outcome of the suit.