Hi
Anthropic to Pay $1.5 Billion+ in Landmark AI Copyright Settlement
In what is being hailed as the largest copyright recovery in U.S. history, Anthropic has agreed to a staggering $1.5 billion settlement—plus interest—to resolve a class action lawsuit brought by authors over unauthorized use of their copyrighted works to train its AI models.
The proposed deal, pending court approval, will see Anthropic pay $3,000 per copyrighted work used without permission, with four payouts scheduled over two years. Payments will go to rightsholders—including authors and publishers—after deductions for legal fees and administrative costs.
According to filings, the current draft includes approximately 465,000 titles. Should the final list exceed 500,000, Anthropic will increase its payout proportionally. The final list of eligible works is expected by October 10, and an official searchable claims database will be created to help authors determine their eligibility. Notice of the settlement and claim-filing instructions are expected to go out this fall.
“This is a game-changing moment,” said Maria Pallante, CEO of the Association of American Publishers (AAP). “It sends a powerful message to every AI company: mining pirated books from illegal shadow libraries is not innovation—it’s infringement.”
Authors and Publishers to Share in Payout
The class action recognizes both legal and beneficial owners of copyrighted works. This means publishers who hold exclusive rights will be compensated, but so will authors whose rights have reverted or whose books are out of print.
To ensure a fair and equitable distribution of funds, a new Author-Publisher Working Group will be formed, co-chaired by Mary Rasenberger, CEO of the Authors Guild, and Pallante. This group will address disputes where multiple parties file claims for the same work.
Anthropic to Purge Pirated Datasets
As part of the agreement, Anthropic will destroy its LibGen and PiLiMi datasets after fulfilling any court-mandated preservation requirements. However, this settlement only covers alleged copyright violations up to August 25, 2025—leaving the door open for future claims, especially those related to infringing outputs from Anthropic’s AI models like Claude.
The unopposed motion for preliminary approval describes the resolution as a “landmark result,” especially given the legal uncertainty. Previously, Anthropic had secured a favorable ruling from Judge William Alsup, who determined that its use of copyrighted materials to train AI models fell under fair use. However, a separate trial was looming over Anthropic’s alleged reliance on pirated ebook libraries, and the judge made clear: “Authors argue Anthropic should have paid for these pirated library copies. This order agrees.”
A Warning Shot to the AI Industry
Rasenberger called out the “criminal-level piracy schemes” allegedly used by Anthropic and other LLM developers, noting that many of the books were sourced knowingly from notorious piracy sites. “This settlement marks a turning point,” she said. “AI developers now know they cannot simply take creative works without consequence.”
What’s Next?
Once the court grants preliminary approval, class members will receive notice with instructions on how to file a claim. Authors Guild emphasized that participation is open to both authors and publishers, with legal ownership and reverted rights determining payment.
This case is likely to set precedent for future copyright actions in the AI space, as creators fight to maintain control over their intellectual property in an age of mass data scraping and machine learning.