Overview
- U.S. District Judge William Alsup declined to approve the proposed class settlement for now, faulting missing details and setting deadlines of September 15 for a finalized works list, September 22 for a claims form, and a follow-up hearing on September 25.
- The deal envisions about $3,000 per eligible title from a $1.5 billion fund, with plaintiffs’ counsel indicating roughly 465,000 works are on the current list.
- Under the proposal Anthropic would destroy the pirated digital files it downloaded but would not delete or alter its trained Claude models.
- The agreement addresses only past conduct through August 25, 2025, reflecting the term sheet’s cutoff and leaving later claims outside the deal.
- Separately, authors Grady Hendrix and Jennifer Roberson filed a new lawsuit against Apple alleging its OpenELM/Apple Intelligence models were trained on pirated books, highlighting expanding legal pressure on AI training data practices.