Federal Judge Approves Historic $1.5 Billion Anthropic Copyright Settlement
The record-breaking deal resolves claims over pirated books used to train the Claude AI model but leaves broader legal questions unanswered.
A federal judge has finalized a historic $1.5 billion settlement between artificial intelligence safety startup Anthropic and a class of authors and publishers, marking a watershed moment in the ongoing legal battles over generative AI training data.
U.S. District Judge Araceli Martínez-Olguín of the U.S. District Court for the Northern District of California signed off on the agreement on Monday. In doing so, the judge rejected several objections to the deal, including claims that the total payout was insufficient, that some copyright holders were unfairly left out, and that the plaintiffs’ legal team was overcompensated. Martínez-Olguín awarded the class attorneys just over $101 million, a significant reduction from the $187.5 million they had initially requested.
The settlement represents the largest financial recovery in the history of U.S. copyright litigation. It addresses allegations that Anthropic ingested approximately 500,000 copyrighted books to train its Claude large language model without permission or compensation.
According to court filings, more than 91% of the eligible authors and publishers represented in the class action have already submitted claims. Under the terms of the agreement, rights holders are expected to receive roughly $3,000 per eligible work before accounting for legal fees, administrative expenses, and other costs. The final payouts will be split between the authors and the publishers holding the respective rights.
Beyond the financial terms, the settlement mandates that Anthropic destroy all copies of the pirated datasets covered by the litigation. However, the agreement does not entirely shield the company from book-related copyright disputes, as several authors and publishers opted out of the class action to pursue independent lawsuits.
The legal battle began in 2024 when authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson filed a class-action lawsuit accusing Anthropic of downloading their books, alongside hundreds of thousands of other publications, from notorious online shadow libraries such as Library Genesis and Pirate Library Mirror. Subsequent court disclosures revealed that Anthropic had compiled an internal training library containing more than seven million pirated volumes.
The litigation took a critical turn in June 2025 under then-U.S. District Judge William Alsup, who issued a split ruling on the boundaries of intellectual property in the AI era. Alsup ruled that using copyrighted texts to train an AI model constitutes “fair use” under U.S. Copyright Office principles because the resulting technology is transformative. However, he drew a sharp distinction regarding data acquisition, ruling that downloading and permanently storing unlicensed, pirated copies of books constitutes a separate and actionable act of copyright infringement.
Alsup also ruled in favor of an unconventional training method utilized by Anthropic, in which the company purchased physical copies of books, sliced off their bindings, scanned the pages into digital format, and subsequently destroyed the physical copies.
By settling the case ahead of a scheduled December 2025 trial, Anthropic avoided the risk of catastrophic statutory damages that legal experts estimated could have reached hundreds of billions of dollars. Because the settlement was reached out of court, the district court’s prior rulings will not be reviewed by an appeals court, meaning the case does not establish a binding nationwide legal precedent for the broader tech industry.
The resolution of this class action represents only a partial relief for Anthropic. The company is currently defending itself against a separate $3 billion lawsuit brought by major music publishers, who allege the AI developer pirated more than 20,000 copyrighted songs to train Claude. Meanwhile, similar high-stakes copyright lawsuits filed by the Authors Guild and various creators against other industry giants, including OpenAI, Meta, and Google, continue to progress through the federal court system.
