(Reuters) – A federal judge in San Francisco ruled late Monday that the use of human books without permission to train artificial intelligence systems is legal under US copyright law.
Overseeing alongside high-tech companies on pivotal questions in the AI industry, US District Judge William Alsp said that humanity will train its large-scale language model by authors Andrea Burtz, Charles Grever and Kirk Wallace Johnson.
However, Alsup also said that the human storage of the author's books in the “Central Library” is in violation of copyright and is not used fairly.
The author's Humanity and Attorneys spokesman did not immediately respond to a request for comment on Tuesday's ruling.
The writers sued humanity last year, claiming that the company supported by Amazon and Alphabet used pirated versions of their books without permission or compensation to teach Claude to respond to human prompts.
Class action lawsuits are one of several brought about by authors, news outlets and other copyright holders against companies such as Openai, Microsoft and the Meta platform over AI training.
Fair use doctrine allows copyrighted works to be used in some circumstances without permission from the copyright owner.
Fair use is an important legal defense for high-tech companies, and the Allsup decision is the first decision to address it in the context of generating AI.
AI companies argue that systems are using copyrighted materials fairly to create new, transformative content and can hum the burgeoning AI industry when forced to pay copyright owners for their work.
Humanity told the court that the book is being used fairly, and that US copyright law “not only encourages AI training, but also encourages, to promote human creativity.” The company said the system copied the book to “study the plaintiff's writings, extract non-copyable information from it, and use what it learned to create innovative technology.”
Copyright holders say AI companies are illegally copying their work and generating competing content that threatens their livelihoods.
Alsup agreed to humanity on Monday, saying its training was “very transformative.”
“Like readers who are aiming to become writers, the LLMS of humanity trained their works to not race first, replicate them or replace them.
(Reporting by Blake Britten in Washington, edited by Chiz Nomiyama and Louise Haven)
