Anthropic owes authors $1.5B for using pirated books - but claiming the money is nearly impossible
Anthropic faces a $1.5 billion settlement with roughly half a million authors after the company downloaded millions of pirated, copyrighted books to train its AI model without permission. The problem: the claims process is broken.
Author Maureen Johnson, who has written 18 books including multiple bestsellers, spent months fighting with Anthropic's settlement website earlier this year. Johnson is owed an estimated $3,000 per book - split 50-50 with her publisher - but the system designed to process her claim was so buggy that basic navigation became a barrier.
A federal judge found last fall that Anthropic's use of the pirated books without author permission violated copyright law, though the judge ruled the use itself constituted fair use. Anthropic admitted it downloaded the pirated material but said it did not use any of those works in publicly released AI technologies.
The settlement covers half a million authors
The class action settlement is the first of its kind against an AI company for training data practices. Similar lawsuits are pending against Meta and OpenAI.
Authors face a fundamental problem: Anthropic's dataset was so poorly organized that the company built a claims website that mirrors those same flaws. Writers trying to verify their books were included in the pirated training set encounter a system that doesn't work reliably.
The settlement represents a rare legal acknowledgment that AI companies cannot simply take published work without permission. But for the authors actually trying to collect, the process has become a test of persistence.
Learn more about AI for Legal issues affecting creators, or explore resources for AI for Writers.
Your membership also unlocks:
