ANI vs. OpenAI dispute: 'Allegations of copyright infringement unfounded'

In the copyright infringement case filed by Asian News International (ANI) against OpenAI, amicus curiae Adarsh Ramanujan argued before the Delhi High Court that the allegations of infringement by the news agency are unfounded.  

Ramanujan asserted that OpenAI’s large language model (LLM) does not reproduce original content but rather predicts language patterns, making it a case of “non-expressive” use.

“The way the model works is that it only processes numbers. Every expression or dataset provided to the program is broken down into individual tokens,” Ramanujan explained in his submission. He further clarified that the model is designed to predict content accuracy without directly reproducing original data, stating, “There is no reproduction of the original data during the training loop.”

During the review of his report, Ramanujan added that infringement could only be established if ANI could prove that ChatGPT’s output contained a direct extract of the “expressive” part of a news article. He also pointed out that the argument currently revolves around ‘fair use,’ whereas Indian copyright law requires an assessment under ‘fair dealing.’

Meanwhile, a separate report submitted by amicus curiae Arul George Scaria informed the court that it has jurisdiction to hear ANI’s copyright suit against OpenAI. Addressing the issue of territorial jurisdiction, Ramanujan argued that ANI should not be compelled to pursue the case in a foreign court simply because the alleged infringement occurred outside India. “If the plaintiff has a right under Indian law, he must have a remedy under Indian law. A foreign court will not have jurisdiction over Indian copyright,” he noted.

ANI countered by asserting that under Indian copyright law, the author of a work is the first owner of the copyright unless the content is created under employment or a contractual agreement for publication.