More than 100 authors sued Anthropic in federal court on June 17, demanding over $75 million for what they call the brazen theft of their books to train AI systems. The complaint, filed in the Northern District of California, escalates the fight over how AI companies obtain the writing that makes their models sound human - and what creators are owed.
The lawsuit names more than 500 pirated titles, including the Oprah-endorsed bestseller "Get Good with Money" by Tiffany Aliche and the international hit "Like Water for Chocolate" by Laura Esquivel. Plaintiffs include Atari co-founder Nolan Bushnell, Academy Award-nominated screenwriter Zachary Sklar, and Newbery Medal winner Donna Barba Higuera. They are asking for $150,000 per infringed work.
How the books were allegedly taken
The authors claim Anthropic used BitTorrent to tap illegal libraries like Library Genesis and Pirate Library Mirror, downloading millions of unauthorized copies. Those texts were then fed into a central repository that trained Anthropic's AI assistant, Claude. Unlike the earlier class action, this suit also accuses the company of redistributing the books internally, giving its models broad access to the copied works.
A separate fight after a $1.5 billion settlement
All 100-plus plaintiffs opted out of a previous class action that settled for $1.5 billion in September. In that case, each author received about $3,000 per stolen work after Anthropic agreed to destroy the downloaded book files. The new filing argues that even if the copies are destroyed, the redistribution and ongoing use of those texts in model training remain unaddressed - and uncompensated.
Lead lawyer James Bartolemei put the core question directly: "Anthropic built the future of artificial intelligence on the books of authors - past, present, and living. My clients ask a simple question: is that innovation, or is it theft with better marketing? Let a jury decide." Anthropic did not return a request for comment.
Why this matters for writers
The per-work damages sought here - $150,000 - set a much higher bar than the $3,000 payout from the earlier settlement. If the redistribution claim holds, it could reshape what "fair use" arguments look like when an AI company copies entire libraries. For writers who depend on copyright to earn a living, these cases will define whether their work becomes training data without consent - or something they can license and control. Following resources on AI for Writers and understanding the technology behind models like Claude can help authors respond as the legal landscape shifts.
Your membership also unlocks:
