Particle.news

Download on the App Store

Anthropic Seeks Court Approval for $1.5 Billion Deal to Settle Authors’ AI-Training Lawsuit

The proposed fund follows a ruling that drew a line between training on lawfully acquired books versus works from pirated libraries.

Overview

  • Filings detail a $1.5 billion compensation pool covering roughly 500,000 books, or about $3,000 per work.
  • The agreement requires Anthropic to destroy pirated documents and copies while preserving rights tied to legally acquired and scanned books.
  • The parties have asked U.S. District Judge William Alsup in San Francisco to approve the settlement, which would avert a December damages trial.
  • Plaintiffs allege the company downloaded millions of books and stored more than seven million pirated titles in a central repository to train Claude.
  • The filing describes the payout as the largest publicly known copyright compensation to date and observers say it could push AI firms toward broader licensing, with other lawsuits still pending.