ANI Media Pvt. Ltd. v. Open AI OpCo LLC

Relevant Acts and provisions

Copyright Act, 1957

Section 14(a)(i)Section 51Section 52Section 52(1)(a)

Provisions considered: Sections 14(a)(i), 51 and 52(1)(a) of the Copyright Act, 1957.

Brief facts

ANI is a news agency that produces and syndicates news articles and interviews. It sued Open AI, the company behind ChatGPT, for copyright infringement. ANI made two claims: first, that Open AI copied and stored ANI's articles to train its language models without permission (the training claim); and second, that ChatGPT reproduced ANI's copyrighted content while answering user prompts (the output claim). ANI sought an interim injunction to stop this use. Open AI's servers are located in the United States. Several industry bodies, publishers and AI companies intervened, some supporting ANI and some supporting Open AI. Two Amici Curiae also assisted the Court.

Issues before the Court

1. Whether Open AI's storing of ANI data for training ChatGPT violates copyright.
2. Is ChatGPT infringing on ANI's copyrighted data by using it to create user responses?
3. Whether Open AI's use of ANI's copyrighted data constitutes fair dealing under Section 52 of the Copyright Act, 1957.
4. Do Indian courts have jurisdiction over the complaint, given Open AI's servers are in the United States?

Court's findings

The Court described that training the model to predict raw data from the internet is divided into tokens and transformed into vectors for providing practical responses. It determined that the training of ChatGPT's models had ceased before ANI's quoted articles were published, so the responses ANI complained about could not have been derived from memorised training data. The Court rejected ANI's argument that only individual humans can claim 'private' or 'research' use under Section 52(1)(a), holding these terms can extend to a company like Open AI, since the training data is never made public and the process itself resembles research. The Court noted that ANI never blocked its website from web crawlers or opted out of scraping, an option available to it, while Open AI had already blocked ANI's website from both training and RAG search. ANI had itself offered Open AI a licence for its content for USD 7.5 million; the Court treated this as proof that ANI's claim was capable of being valued and compensated in money, which weighed against granting an injunction.

Decision

The Delhi High Court dismissed ANI's application for interim injunction. It held, on a prima facie basis, that Open AI's storage of ANI's works for training its LLMs falls under the fair dealing exception in Section 52(1)(a) of the Copyright Act and does not amount to infringement. It also held that ChatGPT's responses were not a substantial reproduction of ANI's works.

Keywords

Artificial Intelligence Copyright Fair Dealing Copyright Infringement Jurisdiction Large Language Model Section 52 Training Data