Judge approves $1.5 billion Anthropic copyright settlement, few authors opt out

A federal judge on Monday confirmed a $1.5 billion settlement that resolves the largest copyright class-action to date involving AI developer Anthropic. The case stemmed from a prior ruling that Anthropic’s use of books for training its models qualified as fair use, while its unauthorized copying of works did not.

Judge Araceli Martínez-Olguín noted that roughly 95 % of the class received notice of the settlement and that about 91 % of affected authors and publishers have already filed claims. Only 350 class members were granted opt-out status, while 54 additional requests were denied, and just two late opt-outs were approved on grounds of excusable neglect.

The court also trimmed the attorneys’ fees and service awards. Lawyers had initially sought 20 % of the fund ($300 million); after negotiations the request fell to 12.5 % ($187 million). Martínez-Olguín ruled that even this amount was excessive, cutting fees to under 7 % of the settlement—approximately $101 million—and ordered a post-distribution accounting that could further reduce fees if warranted.

Lead plaintiffs said the agreement moves the industry toward greater accountability for AI developers. Anthropic’s deputy general counsel, Aparna Sridhar, expressed satisfaction that more than 91 % of right-sholders have claimed their share and that the case’s precedent confirms the company’s training practices as fair use.

The judge emphasized non-monetary benefits of the settlement, including the avoidance of protracted litigation and a requirement that Anthropic destroy the copyrighted works used in training. Any remaining funds after valid claims are paid may be redistributed to class members, unless doing so is economically infeasible.