AI's copyright bill just came due — what the $1.5 billion Anthropic settlement changes
A US court approved the biggest copyright payout in American history against an AI company, setting a precedent for five more AI giants still in court over the same issue.
- $1.5B
- total settlement — largest copyright deal in US history
- $3,100
- paid to each book author
- 480,000+
- books covered by the deal
- 5
- major AI companies still facing similar suits
On Monday, a US judge approved a $1.5 billion settlement between Anthropic and the authors of more than 480,000 books — the largest copyright class action settlement in American history. Anthropic trained its Claude AI on those books, but it had downloaded them from illegal piracy sites, including Library Genesis, a massive archive of stolen texts.
The judge drew a careful line. Training AI on copyrighted material might still be allowed under US law — there is a legal rule called fair use that lets you copy some things without permission. But downloading those books from piracy sites first is a separate act, and that distinction just cost $1.5 billion. Anthropic must also destroy every pirated file it holds.
Individual authors receive roughly $3,100 per book — small for any one person, but historic as a legal precedent. Five major AI companies still face similar suits: OpenAI, Meta, Google, Midjourney, and Perplexity. Each of them is now calculating their risk with this settlement as the reference point.
For anyone building AI products, the message is simple: where your training data comes from is now a legal question, not just an ethical one. Courts have not banned training on copyrighted text — that debate is still going. But getting it from piracy sites is firmly off the table. Licensed data and content that users agreed to share are the safe ground, and knowing that is a genuine business advantage.