Judge Approves Anthropic’s $1.5 Billion Copyright Settlement With Authors
The payout of $3,000 per pirated book across an estimated 500,000 works ranks as the largest copyright recovery on record, yet it leaves the core legal question over AI training unsettled.
WASHINGTON — Anthropic, the San Francisco AI lab behind the Claude chatbot, won final court approval to pay $1.5 billion to authors and publishers who accused it of training its models on pirated books.
The settlement will deliver $3,000 for each of an estimated 500,000 works, shared among the rights holders, and stands as the largest known copyright recovery in U.S. history, according to the law firm representing the plaintiffs. Judge Araceli Martínez-Olguín of the U.S. District Court for the Northern District of California signed the order on July 20, 2026, writing that the deal offers “meaningful relief.”
“We are pleased that more than 91% of authors and publishers covered by the settlement have claimed their share of the payment,” Anthropic general counsel Aparna Sridhar told Reuters. “We’re looking forward to bringing this matter to a close.”
The case reached this point through a split ruling that unsettled creators even as it delivered them a payout. The now-retired Judge William Alsup found that training an AI model on copyrighted text counts as fair use, a conclusion widely read as a turning point for the industry, theverge reported. Alsup separately ruled that Anthropic broke the law by building part of its training library from pirate sites such as Library Genesis and Pirate Library Mirror rather than buying and scanning the books.
Anthropic agreed to settle in September 2025 to avoid a trial on the piracy question and the damages a jury might have imposed, techcrunch reported. Three authors, Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, first filed the copyright suit in 2024.
The payout resolves one dispute without ending the fight. Because Alsup’s decision was a single district court ruling and the settlement blocks any appeal, the fair-use question never reaches a higher court to become binding precedent. Other judges remain free to reach different conclusions, techcrunch reported. Anthropic still faces claims from Chicken Soup for the Soul and other authors who argue $3,000 per book falls short.
Parallel copyright suits continue against Google,Meta, Midjourney, and OpenAI over the same practice. A group of publishers and authors, including Hachette, Cengage, Elsevier, and author Scott Turow, filed a class action against Google over accusations it used copyrighted works to train Gemini, techcrunch reported.
For the average reader, the outcome sets an early price on the books that fuel chatbots millions now use daily, but it does not settle whether such training is legal, leaving the terms of future AI products in flux.
The fight arrives as U.S. AI policy itself splinters. Advisers to Donald J. Trump traded public insults over how to respond to Kimi, a free open-weight model from Chinese company Moonshot that rivals paid U.S. systems, technologyreview reported. Moonshot plans to release the full model for free download on July 27, 2026, according to fisherphillips.com, and federal officials are weighing export-style restrictions on Chinese models that stop short of an outright ban.
Has an AI copyright dispute settled at this scale before? It has not. The prior benchmarks in AI litigation were smaller and mostly unresolved, and no earlier case had put a defined dollar figure on pirated training data at this magnitude.
Remaining Anthropic claimants and the judges overseeing the Google,Meta, Midjourney, and OpenAI suits will decide the next moves, though no dates for those rulings have been announced.
Copyright law has bent to new technology before. When Napster popularized file-sharing in 1999, the recording industry sued, and a federal court ordered the service shut down in 2001 after finding it liable for enabling mass infringement. The label victory reshaped digital music but did not stop copying; it pushed the industry toward licensed platforms and, eventually, streaming. Anthropic’s settlement echoes that pattern, drawing a line under one company’s conduct while the larger question of how creative work feeds machines stays open.