Judge approves Anthropic’s $1.5 billion author settlement

A federal judge approved Anthropic’s $1.5 billion settlement with authors and publishers, resolving a copyright class action over books used in connection with AI development. The order matters for AI companies because it leaves intact a ruling favorable to model training while forcing a large payment tied to alleged piracy of copyrighted works.

U.S. District Judge Araceli Martínez-Olguín approved the settlement Monday, rejecting objections from authors who argued that the deal paid too little and that legal fees were too high. The case followed a court ruling that Anthropic’s use of books to train AI was fair use, while its alleged acquisition of pirated works was likely unlawful.

The settlement covers 506,194 works, according to the court order. The expected payout is about $3,000 per work, which Martínez-Olguín noted is four times the minimum statutory damages. The final amount paid per work could rise if money remains after valid claims are paid and redistribution is economically feasible, the judge said.

Participation was high. Martínez-Olguín said about 95 percent of the class received notice, and about 91 percent of affected authors and publishers had filed claims. Only 350 class members opted out. Another 54 either objected or tried to opt out after the deadline.

That claim rate helped the court reject arguments that the settlement process was unfair. Martínez-Olguín wrote that the response indicated most class members received timely notice, considered the distribution plan fair and supported the deal.

Fees reduced from requested levels

The judge approved the headline settlement amount but cut the lawyers’ fee award. Plaintiffs’ lawyers initially sought 20 percent of the fund, or $300 million, before lowering the request to 12.5 percent, about $187 million. Martínez-Olguín reduced the award to less than 7 percent, or about $101 million.

Some authors had challenged the fee request, including projected work tied to distributing payments. The judge said the court would require a post-distribution accounting after payouts are completed and could reduce fees if the accounting supports doing so.

The court also reduced service awards for the three named author plaintiffs. They had sought $50,000 each. Martínez-Olguín approved $15,000 each, finding the higher figure unreasonable despite the time and resources they spent on the case and settlement negotiations.

The lead plaintiffs told Reuters the settlement moves the case toward accountability for Anthropic and warns AI companies against bypassing creators’ rights. Plaintiffs’ lawyers said in a press release that rightsholders had shown broad support for the settlement.

Anthropic’s deputy general counsel, Aparna Sridhar, said in a statement to Ars Technica that the company was pleased that more than 91 percent of covered authors and publishers had claimed a share of the payment and that Anthropic was looking to close the matter. She also pointed to the court’s fair-use ruling on AI training.

Most late opt-out requests failed

Anthropic opposed delayed opt-outs, which could expose the company to separate lawsuits. Martínez-Olguín denied most late requests, granting only two where authors showed excusable neglect. One involved a co-author who filed a few days late after not receiving notice. Another involved a co-author who had suffered a stroke, lived in Mexico, spoke Spanish and said no Spanish translation was provided.

The judge denied a late request from entertainment lawyer and author Donald Passman, who sought to opt out at the end of June, three months after the March 30 deadline. Passman argued in a court filing that he did not receive timely notice and that the estimated payment for his book, All You Need to Know About the Music Business, did not adequately compensate him for potential individual copyright claims.

Anthropic argued that allowing Passman’s late opt-out would invite similar requests and delay distribution. Martínez-Olguín agreed, finding the request inexcusably late and warning that granting it could encourage more authors to try to leave the settlement.

The settlement also includes non-cash terms. Martínez-Olguín said it lets class members avoid costly litigation that could end with no recovery, requires Anthropic to destroy the covered works, and does not block future litigation if Anthropic later misuses those works.

This story draws on original reporting from Ars Technica.