Judge Decides Anthropic AI Training On Sah Copyright Work, Still Use Is Fair

Photo: Anthropic

JAKARTA - Anthropic appears to have won over allegations of copyright infringement filed by several authors regarding the use of their work for training his artificial intelligence model (AI).

Federal judges in San Francisco decided that the use of books by Anthropic without permission to train its AI system was legitimate under US copyright law. US District Judge William Alsup even said Anthropic was doing "a reasonable use".

"This order provides a summary decision for Anthropic that the use of training is a reasonable use," Alsup wrote in his ruling.

However, the company is still caught in a legal case and must stand trial for the way they obtained the books, which are claimed to download books from online shadow library' containing pirated books.

Alsup also stated in his ruling that Anthropic should still be tried in December for alleged theft of their work to train his Claude chatbot.

"We will conduct a trial of a pirated copy used to create an Anthropic central library and the harm it causes, both actual and under law," the ruling continued.

He also emphasized that if Anthropic bought an official copy of the books previously illegally retrieved from the internet, it would not be a move that would immediately free the company from legal responsibility.

However, this can be a mitigating factor in determining the amount of compensation that must be paid.

The case began in late 2023, where Joseph Bartz, author and educator, along with a number of large publishers, sued the AI Anthropic company in California's US District Court, with the use of pirated copies of copyrighted books to train the AI model Claude, without permission or compensation.

The plaintiff accused of intentional and large-scale copyright infringement, and demanded compensation for the losses incurred.