Anthropic's $1.5 billion copyright settlement with book authors is on hold after US District Judge Araceli Martinez-Olguin declined to grant final approval on May 15, 2026, citing objections from class members who argue the deal hands $320 million to plaintiffs' lawyers while paying individual authors roughly $3,000 each. The settlement, billed as the largest copyright resolution in US history, covers more than 480,000 works that Anthropic obtained through what the underlying complaint described as wholesale book piracy used to train its models. Authors and copyright holders have filed claims on over 92% of those works, according to an attorney for the plaintiffs.
Martinez-Olguin asked authors' counsel to respond by May 21 to objections she said warranted scrutiny before she signs off. Anthropic has been ordered to file a brief on the same deadline explaining why late opt-outs should not be honored. A separate group of 25 class members has already opted out and filed a new lawsuit against the company, signaling that the legal exposure does not end with this settlement even if it clears the court.
The loudest objection targets the fee structure. Pierce Story, an objector with two works in the class, estimated that the $320 million request works out to roughly $10,000 to $12,000 per hour for the plaintiffs' lawyers, including a generous projection of future hours. Story cited an 8th Circuit ruling in a T-Mobile case where the court observed that "no reasonable class member would willingly pay" a fee in the lower range of $7,000 to $9,500 per hour.
Key facts
- 01Judge Araceli Martinez-Olguin declined to approve Anthropic's $1.5B settlement on May 15, 2026, citing objections over attorney fees.
- 02Plaintiffs' counsel requested $320M in legal fees while individual authors expect roughly $3,000 per claim.
- 03Authors filed claims on over 92% of the 480,000 works covered, while 25 class members opted out and filed a separate lawsuit.
- 04Objector Pierce Story proposed cutting counsel fees to $70M, which would raise individual author awards by about 25%.
- 05Anthropic and authors' lawyers must respond to objections by May 21.
Story is asking the court to recalibrate the split. "Every dollar that Counsel takes from the Settlement fund is one that is not given to those actually harmed," he wrote. His proposed alternative would cap counsel fees at $70 million, which he calculates would deliver a roughly 25% bump to individual author awards while still paying the lawyers at their stated top rates for hours worked.
“Story estimates the legal fees translate to $10,000–$12,000 per hour, citing an 8th Circuit ruling that found no reasonable class member would pay even $7,000–$9,500.”— Jaeden Schafer
He also accused the legal team of breaking an earlier promise to tie their compensation to actual member payouts rather than the gross settlement amount. "Were the attorneys as skilled, gritty, and brilliant as they profess, and were the Settlement the 'home run' Counsel claims it to be, Plaintiffs would receive more than this pittance," Story wrote in his filing. Another objector, Ruben Lee, echoed the point: "I believe the amount offered is paltry, and does not in any way reflect the full value of the unauthorized use of my work."
A separate strand of objections focuses on what Anthropic gets to keep doing after the settlement closes. James R. Sills, who has two works in the class, argued that Anthropic should be required to destroy all scanned copies of the works, both digital and physical, before the deal proceeds. "Currently, Anthropic will not delete any scanned physical copies of works/books," Sills wrote. "So, they currently can use these works."
Robert C. Jacobson raised a related concern about "the absence of any prospective relief or framework addressing ongoing commercial use of models trained on the class works." Models that have already ingested the pirated material remain in production and continue to generate revenue, a fact the settlement does not directly address. For objectors, paying $3,000 per author without restricting the downstream use of the trained weights leaves the underlying business intact.
Several objectors also described procedural friction in getting their complaints on the record. Lee said he "tried to file this objection via the Court's ECF and PACER systems, but have found it impossible to do so." Victoria Pinder said the authors' legal team initially docketed her objection, then later tried to mark it invalid by claiming it had never been submitted, and misspelled her name in the filing that attempted to exclude her. Jacobson said counsel characterized his objection as improperly filed despite the court acknowledging delays in docketing earlier submissions.
The objections have a sympathetic precedent inside the case. Judge William Alsup, who initially approved the settlement before retiring, had questioned whether the legal fees were excessive and warned that the deal risked being "shoved down the throat of authors." According to objector Lea Bishop, a copyright law professor who is not a class member, Alsup recommended an independent investigation into the fee request, but that recommendation was not squarely disclosed to Martinez-Olguin in the status report submitted by plaintiffs' counsel, and class members were not notified of it.
Anthropic and the lead plaintiffs' lawyers have not yet filed their May 21 responses, and the company has consistently maintained that the settlement resolves the claims covered by the class. Whether Martinez-Olguin reduces the fee award, restructures the payout formula, or pushes for additional restrictions on Anthropic's use of the underlying works remains open. Story warned in his filing that approving the deal as written would be "an aberration of civil justice and a slap in the face to all those who labored to publish their works."
The fight matters well beyond this docket. Anthropic's settlement is the template every other AI lab is watching as copyright cases against OpenAI, Meta, and Google grind forward, and the per-work economics being set here will shape the negotiating floor for the next wave. If Martinez-Olguin trims counsel fees or forces concessions on model deletion and prospective use, the cost of training on copyrighted text rises for every lab still doing it, and the settlements get harder to close quietly. If she approves the deal as filed, the message to AI companies is that $1.5 billion buys closure on 480,000 books with no operational changes — a number that, divided across the industry, starts to look like a manageable line item rather than a deterrent.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
