A federal judge has approved Anthropic’s $1.5 billion settlement with authors and publishers, closing a copyright class action over books used in connection with the company’s AI systems and leaving only a small group of rightsholders outside the deal.
US District Judge Araceli Martínez-Olguín overruled objections to the settlement on Monday, finding that the challenges lacked merit. According to her order, roughly 95 percent of the class received notice, and about 91 percent of affected authors and publishers have submitted claims. The court said 350 class members opted out, while 54 others objected or tried to leave the class after the deadline.
The case followed an earlier ruling that Anthropic’s use of books to train AI qualified as fair use, while the company’s alleged acquisition of pirated copies likely did not. That distinction is the whole machine room of the dispute: training was not the legal problem the court saw, but the route Anthropic allegedly took to get some of the books was.
The settlement is expected to pay about $3,000 per work, an amount Martínez-Olguín said was four times the minimum statutory damages. Some authors argued that was too little and that copyright law could support higher awards in separate lawsuits. The judge was not persuaded that the class deal should be rejected on that basis.
Fees cut sharply
Martínez-Olguín approved the settlement fund but reduced what the lawyers and named plaintiffs will receive. Class counsel initially sought 20 percent of the fund, or $300 million. Before the ruling, the request was lowered to 12.5 percent, about $187 million.
The judge still found that request too high. She cut attorneys’ fees to under 7 percent of the settlement fund, about $101 million. Because some of that money covers future work distributing payments, Martínez-Olguín said the court will require a post-distribution accounting and may reduce the fees further if the actual work does not support them.
The three authors who represented the class also received less than they requested. They sought service awards of $50,000 each. Martínez-Olguín reduced those awards to $15,000, finding the larger amount unreasonable even though the named plaintiffs spent substantial time and resources on the case and settlement talks.
Lead plaintiffs told Reuters that the settlement “brings us one step closer to real accountability for Anthropic and puts all AI companies on notice they can’t shortcut the law or override creators’ rights.”
Anthropic’s deputy general counsel, Aparna Sridhar, said in a statement to Ars Technica that the company was pleased with participation in the claims process. “We are pleased that more than 91 percent of authors and publishers covered by the settlement have claimed their share of the payment, and we’re looking forward to bringing this matter to a close,” Sridhar said.
Late opt-outs mostly failed
Anthropic opposed late opt-out requests, which would have preserved the possibility of separate lawsuits. Martínez-Olguín accepted only two late requests from authors who showed excusable neglect. One was filed a few days late by a co-author who did not receive notice. Another was granted after a longer delay involving an author who had suffered a stroke, lived in Mexico, spoke Spanish and said no Spanish translation of the notice was provided.
The judge rejected other late attempts, including one by Donald Passman, an entertainment lawyer and author of All You Need to Know About the Music Business. Passman argued in a filing that he did not receive timely notice and that the estimated payment did not compensate him adequately for the potential value of his individual claim. Martínez-Olguín found his request, submitted about three months after the March 30 deadline, inexcusably late.
Anthropic argued that allowing Passman to leave the settlement that late would encourage others to try the same tactic and delay payments. Martínez-Olguín agreed that granting the request risked opening the door to more late exits.
The judge also pointed to non-cash terms in the settlement. According to her order, the deal spares class members from costly litigation that could have produced no recovery, requires Anthropic to destroy covered works, and does not bar future litigation if Anthropic misuses works later.
Authors may receive more than the estimated per-work amount if money remains after valid claims are paid, Martínez-Olguín said. Any redistribution would depend on whether doing so is economically feasible.
This story draws on original reporting from Ars Technica.
