Judge signs off on Anthropic’s $1.5 billion authors settlement
A San Francisco federal judge approved a deal over pirated books used by Anthropic, with eligible authors and publishers set to claim about $3,000 per work.
By Georgia Hale · Staff Writer
3 min read
A federal judge in San Francisco has given final approval to Anthropic’s $1.5 billion settlement with authors who accused the AI company of illegally downloading copyrighted books, ending a case the court described as the largest copyright class action settlement in history.
U.S. District Judge Araceli Martínez-Olguín signed the approval order on July 20, resolving claims from lead plaintiffs Andrea Bartz, Kirk Wallace Johnson and others. The authors alleged that Anthropic used pirated libraries, including LibGen and PiLiMi, to assemble a collection of books used in connection with its AI work.
The fight focused on how Anthropic got the books. According to the court’s order, the settlement does not decide broad questions about whether copyrighted material can be used to train AI systems.
That issue had already been addressed in an earlier ruling cited by the court. In that decision, the court found that the act of training on the material qualified as fair use.
What authors can claim
Under the settlement, authors and publishers whose pirated books appear on Anthropic’s “Works List” can seek roughly $3,000 for each book, according to the court order. The order said that figure is about four times the usual statutory minimum in copyright infringement cases.
More than 91 percent of eligible works have already been claimed, covering over 440,000 books, according to the order.
Anthropic must also delete the pirated files it downloaded. The settlement releases the company from liability tied to its past acquisition of that training data, according to the court, but it does not shield Anthropic from every possible AI-related claim.
Judge Martínez-Olguín said the deal does not release claims for future harm or claims “based on the output of AI models.” That leaves the door open for later disputes over what Anthropic’s systems produce, or over new conduct after the covered period.
Objections rejected
The court overruled all 54 objections and comments filed by class members and third parties, according to the order.
Some objectors wanted the settlement to cover more works. Others asked for non-cash remedies, including attribution for source material or deletion of Anthropic’s models. The court found those requests went beyond the claims this lawsuit could resolve.
With final approval granted, the case is closed. The court said it will continue to oversee the administration of the settlement, including how funds are distributed to authors and publishers whose works qualify.
The ruling arrives as AI companies face growing scrutiny over the data used to build and train their systems. In this case, the court’s final order keeps the spotlight on acquisition of copyrighted works, while leaving future fights over AI outputs and later conduct for another day.
This story draws on original reporting from Mashable.