San Francisco, CA — A federal judge ruled Monday that copyrighted books can be utilized to train artificial intelligence models without the authors’ consent, marking a significant legal precedent for the tech industry.
U.S. District Judge William Alsup’s decision favored Anthropic, a San Francisco-based startup co-founded by former OpenAI employees and backed by Amazon. The ruling comes after the company faced a lawsuit from authors Andrea Bartz, Charles Graeber, and Kirk Wallace, who claimed their works were used without authorization in the training of Anthropic’s AI assistant, Claude.
In his ruling, Alsup stated that Anthropic’s use of purchased books was “exceedingly transformative and was a fair use.” However, he also noted that the company may have violated copyright law by using pirated copies of books. A trial is scheduled for December to address potential damages, which can total as much as $150,000 for each instance of willful infringement.
The judge emphasized that using copyrighted materials to inform AI training could be seen as a transformative process. He questioned whether a person reading classic literature and subsequently emulating various writing styles would constitute a copyright violation, concluding it would not.
Alsup highlighted that Anthropic had unlawfully accessed over 7 million books from online repositories known for housing unauthorized copies. As the company transitioned away from pirating for legal reasons, it hired Tom Turvey, a former Google employee, to manage the procurement of books while navigating commercial and legal challenges.
Initially, Turvey explored licensing agreements with major publishers but ultimately opted to buy millions of print copies in bulk. According to court documents, the company then dismantled these books, scanning their pages into digital formats for machine processing.
While some authors raised concerns over Anthropic’s digitization practices, Alsup ruled that converting physical books to digital files served a transformative purpose by saving space and enhancing searchability. He noted that while Anthropic’s later actions to purchase books might not absolve the company of liability, it could influence the magnitude of potential statutory damages.
This ruling comes amid ongoing litigation involving major studios like Walt Disney Co. and Universal Pictures, which are taking action against artificial intelligence company Midjourney. The studios allege that Midjourney trained its image generation algorithms using copyrighted materials from their films, and the outcomes of these cases could establish significant precedents in the realm of artificial intelligence and copyright law.
This article was automatically written by Open AI, and the people, facts, circumstances, and story may be inaccurate. Any article can be requested to be removed, retracted, or corrected by writing an email to contact@publiclawlibrary.org.