Marking another twist in one of the AI industry’s most closely watched copyright cases, a federal judge has delayed final approval of Anthropic’s proposed $1.5 billion copyright settlement with authors who accused the AI company of illegally using pirated books to train Claude,

At a hearing in San Francisco on May 14, 2026, U.S. District Judge Araceli Martinez asked both sides for more information, particularly regarding attorneys’ fees and payments to lead plaintiffs.

The proposed deal covers 482,460 eligible works and is described as the largest known U.S. copyright settlement. Claims were filed for 440,490 of those works, which is about 92%, against a typical 10% in class action cases. Each eligible work is expected to receive approximately $3,000 after fees and costs, according to the Authors Guild.

The legal battle began after a group of writers sued Anthropic, backed by Amazon and Alphabet, in 2024, accusing the company of downloading pirated books from Library Genesis and Pirate Library Mirror to train Claude without permission.

In June 2025, now-retired Judge William Alsup ruled that training on legally acquired books constituted fair use but declined to extend that finding to piracy, leaving Anthropic exposed to a December trial on damages before the settlement was reached.

Objections at the hearing included undercounting of group copyright registrations, exclusion of foreign works not registered with the U.S. Copyright Office, and disputed publisher-author splits. One class member said a publisher claimed 88% of her per-work payout, leaving her 12%, according to Authors Alliance, an authors’ advocacy organization.

Separately, 28 authors who opted out, including Dave Eggers and Vendela Vida, filed a new copyright infringement lawsuit against Anthropic in California on Wednesday, requesting a jury trial.

Also Read: Anthropic Wins as Judge Temporarily Blocks Trump Administration Ban in DoD Dispute