ANI vs OpenAI: Delhi High Court refuses immediate stay on copyright ruling

On Tuesday, the Delhi High Court declined to grant an immediate stay on a recent copyright ruling that permits the use of copyrighted works for training large language models (LLMs). The news agency ANI had appealed against a July verdict, which allowed developers like OpenAI to continue training on Indian-origin content without licensing agreements.

The division bench of Justices Avneesh Jhingan and Manmeet Pritam Singh Arora suggested that the court would prefer to hear both sides before making a judgment, setting the next hearing for December 8. ANI initiated legal proceedings in 2024 against OpenAI Inc and OpenAI OpCo LLC, the first of its kind in India, seeking an injunction against the use of its copyrighted content in LLM training.

The single judge had initially denied granting any interim injunction, even while the case proceeded. The court ruled that using ANI's original literary works to train LLMs would be considered private use, including research, thus qualifying as fair use. ANI had also objected to the verbatim replication of its copyrighted material when queried in ChatGPT and the hallucinatory responses from ChatGPT falsely attributed to ANI.

In its appeal, ANI argued that fair dealing under the Copyright Act is a specific exception for private or personal use, and that commercial exploitation at scale cannot be inferred from such use. They emphasized that public interest and AI growth should not override statutory rights and criticized OpenAI's policy of excluding website owners from opting out of scraping for LLM training.

OpenAI countered that after the pre-training phase, their models no longer have access to the original training data, and they had previously "blocklisted" ANI's domain to exclude it from future training.