Authors, publishers near final approval of $1.5 billion Anthropic copyright settlement

SAN FRANCISCO (CN) — A $1.5 billion settlement over claims AI developer Anthropic pirated hundreds of thousands of copyrighted books to train its large language models inched toward final approval Thursday.

Plaintiffs, owners of the copyrighted works that Anthropic pirated, sued the company in 2024, arguing that the AI company knowingly trained its large language models on datasets that included massive amounts of copyrighted works without obtaining permission or compensating the authors.

The plaintiffs say Anthropic trained its AI models on datasets it knew contained hundreds of thousands of copyrighted books from websites known to have pirated material, such as Library Genesis and Pirate Library Mirror. They claimed Anthropic’s mass use of pirated material violated copyright law and deprived authors of book sales and licensing revenues, all while building a profitable generative AI tool that could threaten writers’ careers.

“It is not consistent with core human values or the public benefit to download hundreds of thousands of books from a known illegal source,” the plaintiffs say in their complaint. “Anthropic has attempted to steal the fire of Prometheus. It is no exaggeration to say that Anthropic’s model seeks to profit from strip-mining the human expression and ingenuity behind each one of those works.”

The parties were heading to a trial in December 2025 before reaching an agreement in late August. Former Senior U.S. District Judge William Alsup, a Bill Clinton appointee, granted preliminary approval on Sept. 25, 2025.

“I don’t see how you can get a better deal,” he said at the time.

Under the settlement terms, Anthropic agreed to pay $1.5 billion, plus interest, that would be allocated to class members proportionally based on their number of eligible works.

According to the plaintiffs, approximately 482,000 works are included in the settlement, with each class member estimated to receive $3,100 per work, according to plaintiffs’ attorney Justin Nelson of Susman Godfrey.

Anthropic already paid $300 million to the fund shortly after receiving preliminary approval of the settlement, with an additional $300 million expected five days after final approval. The additional payments will be split into $450 million on both the first and second anniversaries of preliminary approval.

In addition to monetary relief for the plaintiffs, Anthropic will also be required to destroy the original files of the works they downloaded from the pirated book datasets within 30 days of the final judgment and certify that Library Genesis and Pirate Library Mirror datasets with pirated material were used in training any of the company’s commercially released large language models.

Class members in the case include “all beneficial or legal copyright owners of the exclusive right to reproduce copies of any book in the versions of LibGen or PiLiMi downloaded by Anthropic. “

Books included in the class are any work “possessing an ISBN or ASIN which was registered with the United States Copyright Office within five years of the work’s publication and which was registered with the United States Copyright Office before being downloaded by Anthropic, or within three months of publication.” Notably, non-US-registered works are excluded from the scope of the settlement.

More than 506,000 potential class members representing approximately 480,000 works, or 99.5% of the works list, received direct notice of the settlement. According to Nelson, nearly 93% of the class has submitted a claim, covering approximately 448,000 works. An additional 350 class members have opted out of the settlement, and 53 have objected.

At Thursday’s hearing, Nelson said the reasons to approve the settlement “have only gotten stronger,” adding that many of the objections to the settlement are from class members asking for more of their work to be added to the eligible list, indicating that the agreement is beneficial for class members.

Judge Araceli Martínez-Olguín, a Joe Biden appointee, allowed objectors to address the court, where several spoke about the concerns they had with how the plaintiffs put together the eligible works list.

One class member told the judge that the works list undercounts the number of eligible works in the class by treating each copyright registration number as a single work, regardless of how many books are covered by the registration. The class member explained that she has certain group copyright registration numbers that include 40 separate, independently published novels under one registration number, all of which were downloaded by Anthropic without permission. However, under the current terms of the settlement agreement, the novels would be considered just one claimable work.

Another class member spoke to the exclusion of works that were published under a pseudonym, disadvantageous to small publishers and self-published authors in the class, while a third said they believed a one-time payment was not enough because Anthropic was continuing to profit off the copyrighted work they stole.

James H. Bartolomei III of Duncan Firm, an attorney representing four other objectors, asked the court to reopen the opt-out period as certain key documents from the case were only uploaded to the settlement website recently.

“Nothing I am asking for takes a dollar away from any class member who filed a claim. I’m not asking the court to stop the settlement from ever being approved. Just for sufficient information to make an informed choice,” he said.

The plaintiffs requested attorneys’ fees of 20% of the settlement, or approximately $300 million. They additionally asked for nearly $2 million in litigation expenses, $17 million in a reserve cost fund for future expenses and $50,000 for each of the three settlement class representatives.

Martínez-Olguín said she would take the final approval motion under submission and ask for a brief submission from the parties to “help keep this moving.”

Representatives for the plaintiffs declined to comment.

A representative for Anthropic did not immediately respond to a request for comment.

Follow @mhattridge_