US court approves Anthropic’s historic $1.5B settlement with authors over pirated training books, paying $3,100 per registered work.
In what marks the largest copyright recovery in United States history, the legal boundaries surrounding artificial intelligence training datasets have reached a watershed financial precedent. On Monday, July 20, 2026, a federal judge granted final approval to a $1.5 billion class-action settlement between artificial intelligence startup Anthropic and thousands of authors and book publishers. The landmark judicial order officially brings closure to the high-stakes lawsuit Bartz v. Anthropic, resolving allegations that the San Francisco-based AI developer unlawfully acquired massive repositories of pirated books to train its flagship Claude large language model.
The technical core of this historic ruling centers on dataset hygiene and intellectual property acquisition rather than the mechanics of machine learning itself. The class-action litigation, originally initiated in 2024 by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, accused Anthropic of downloading more than seven million copyrighted works from known shadow libraries such as Library Genesis and Pirate Library Mirror. Under the approved terms of the $1.5 billion payout pool, rightsholders covering approximately 480,000 registered titles will receive payouts estimated at $3,100 per eligible work. In addition to monetary compensation, the settlement mandates that Anthropic permanently delete and destroy all original files, duplicates, and database mirrors obtained from unverified pirated repositories.
The geographic and jurisdictional anchor of this decision was formalized as sitting inside the U.S. District Court for the Northern District of California in San Francisco. U.S. District Judge Araceli Martínez-Olguín issued the final sign-off, rejecting final objections from a minority of authors who argued the payout failed to reflect potential statutory damages that could have reached hundreds of billions of dollars at trial. Judge Martínez-Olguín noted that criticism of the payout size lacked a realistic assessment of trial risks, approving $101 million in legal fees for the plaintiffs’ counsel team.
Tracing the timeline of this legal battle highlights a rapid two-year pivot from existential trial risk to formal resolution. Preliminary approval for the settlement was granted in September 2025 by former Senior U.S. District Judge William Alsup before his retirement. Following intense claims administration over late 2025 and early 2026, court filings confirmed that rights holders submitted claims covering over 92% of all eligible titles in the class. The July 20, 2026 final approval order clears the path for immediate payout distributions to class members.
See Also: Millions of WordPress Sites at Risk as Hackers Weaponize “WP2Shell” Core Exploit Chain
Understanding the fundamental behind Anthropic’s willingness to write a $1.5 billion check reveals a critical legal distinction established during pre-trial motions. During earlier summary judgment proceedings, Judge Alsup ruled that using copyrighted text to train AI models constitutes fair use because generative outputs are inherently transformative. However, the judge held that downloading millions of pirated files from illicit torrent hubs to build a permanent internal library was an independent act of copyright infringement. By settling, Anthropic insulated itself from catastrophic statutory piracy penalties while preserving the favorable fair-use finding regarding the training process.
About the Author
Jennifer Sakmufuwo Baba is a tech analyst and writer covering artificial intelligence, fintech, and emerging technologies at TechRegard. Based in Nigeria, she's passionate about translating complex tech developments into compelling, accessible stories for diverse audiences. Her work focuses on how technology shapes innovation across Africa and globally.
