FILE PHOTO: Anthropic logo is seen in this illustration taken May 20, 2024. REUTERS/Dado Ruvic/Illustration/File Photo
(Reuters) -A California federal judge ruled on Thursday that three authors suing artificial intelligence startup Anthropic for copyright infringement can represent writers nationwide whose books Anthropic allegedly pirated to train its AI system.
U.S. District Judge William Alsup said the authors can bring a class action on behalf of all U.S. writers whose works Anthropic allegedly downloaded from "pirate libraries" LibGen and PiLiMi to create a repository of millions of books in 2021 and 2022.
Alsup said Anthropic may have illegally downloaded as many as 7 million books from the pirate websites, which could make it liable for billions of dollars in damages if the authors' case is successful.
An Anthropic spokesperson said the company was considering options to challenge the ruling and that the court failed to account for the difficulty of establishing copyright ownership "millions of times over in a single lawsuit." An attorney for the authors declined to comment on the decision.
Andrea Bartz, Charles Graeber and Kirk Wallace Johnson sued Anthropic last year, arguing that the Amazon- and Alphabet-backed startup used their books without permission or compensation to teach its chatbot Claude to respond to human prompts.
The case is one of several high-stakes lawsuits brought by authors, news outlets and other copyright owners against companies including OpenAI, Microsoft and Meta Platforms over their AI training.
AI companies argue their systems make fair use of copyrighted material to create new, transformative content. Alsup determined in June that Anthropic's AI training made fair use of authors' works, but said the company still violated their rights by saving pirated copies of their books to a "central library of all the books in the world" that would not necessarily be used for AI training.
Alsup said on Thursday the three authors could represent all writers whose books Anthropic allegedly downloaded from LibGen and PiLiMi, rejecting Anthropic's argument that identifying all of the copyright-eligible works and their authors would be impractical.
(Reporting by Blake Brittain in Washington; Editing by David Bario, Rod Nickel and Richard Chang)
