Judge allows authors to form class action in Anthropic book piracy lawsuit

U.S. District Judge William Alsup has certified a class action lawsuit that allows authors nationwide to collectively sue AI company Anthropic for allegedly downloading up to 7 million books from pirate websites to train its Claude AI assistant, potentially exposing the company to billions of dollars in damages if the case succeeds at trial in December 2025.

Judge William Alsup's Ruling

On June 23, 2025, Judge William Alsup issued a nuanced ruling in the Bartz v. Anthropic case that both sides could claim as a partial victory. He determined that using copyrighted books to train AI models constitutes fair use, calling it "quintessentially transformative." However, he ruled against Anthropic regarding their acquisition of "pirated" copies for their "central library," finding that downloading millions of books from unauthorized websites without payment was not protected by fair use.whitecase+2

The ruling draws a critical distinction between the legality of training AI on copyrighted works versus building a permanent library of pirated content. While Judge Alsup granted Anthropic's motion for summary judgment on the training issue, he ordered that damages for the pirated copies used in Anthropic's library will proceed to trial. This landmark decision represents the first judicial interpretation of how fair use doctrine applies to AI training data, potentially influencing dozens of similar lawsuits against companies like OpenAI, Google, and Meta.goodwinlaw+1

7 Million Pirated Books

Anthropic's acquisition of copyrighted materials was far more extensive than initially reported. Court findings revealed that co-founder Ben Mann systematically downloaded pirated content from multiple sources: 196,640 books from Books3 in early 2021, approximately five million books from Library Genesis (LibGen) in June 2021, and an additional two million books from the Pirate Library Mirror (PiLiMi) in July 2022. Internal communications showed that despite growing "legal concerns" about using pirated books, Anthropic executives, including CEO Dario Amodei, deliberately chose this path to avoid what Amodei characterized as "legal/practice/business slog."publishersweekly+1

Beyond digital piracy, Anthropic also purchased millions of physical books, which were then "destructively scanned" – their bindings cut and pages sliced for digitization before the original books were discarded. These activities contributed to what the court described as Anthropic's attempt to create a "central library of all the books in the world." This comprehensive collection of literary works was deemed crucial by the company, which had "become convinced that using books was the most cost-effective means to achieve a world-class LLM" capable of mimicking high-quality human writing.hackernoon+4

LibGen and PiLiMi Libraries

Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi) are among the most notorious platforms for accessing pirated books that played a central role in Anthropic's training data acquisition. These websites collectively host millions of unauthorized digital texts—LibGen contains over 7.5 million books and 81 million research papers—operating in legal gray areas where copyright enforcement is typically lax. Their existence highlights the tension between democratizing knowledge access and protecting intellectual property rights, with AI companies like Anthropic, Meta, and OpenAI exploiting these repositories for commercial gain.publishersweekly+1

The importance of these pirated libraries to AI development was underscored in court documents, which revealed that Meta executives considered books "more important than web data" for AI training, eventually securing permission from CEO Mark Zuckerberg to download and use LibGen's dataset. Similarly, Anthropic co-founder Ben Mann deliberately targeted these platforms, downloading five million books from LibGen in June 2021 and two million more from PiLiMi in July 2022. This deliberate exploitation of pirated content for commercial AI development has sparked numerous copyright lawsuits, creating significant legal exposure for companies that have utilized these unauthorized sources.transparencycoalition+3

July 17 class action certification

On July 17, 2025, Judge William Alsup certified a nationwide class action against Anthropic, dramatically transforming the legal landscape of the case. The certification allows authors whose works were allegedly downloaded from LibGen and PiLiMi to collectively sue Anthropic, rather than pursuing individual claims. The judge rejected Anthropic's argument that identifying all copyright-eligible works and their authors would be impractical, though he denied certification for authors whose works appeared in the Books3 dataset or were legitimately purchased and scanned by the company.winbuzzer+2

The class action certification significantly raises the financial stakes for Anthropic, potentially exposing the company to business-ending liability. With statutory damages ranging from $750 to $30,000 per work infringed (up to $150,000 for willful infringement), even modest damages across thousands of works could result in billions in liability. Legal experts note this represents a "critical step" forward for authors, as Anthropic must now defend against a unified front of potentially hundreds of thousands of writers whose works were allegedly pirated. The court appointed Lieff Cabraser Heimann & Bernstein LLP and Susman Godfrey LLP as lead class counsel, with a trial scheduled to begin December 1, 2025.law360+3