Anthropic just found out what pirated books cost once you get caught: $1.5 billion, the largest copyright settlement in U.S. history.
U.S. District Judge Araceli Martinez-Olguin granted final approval on Monday to Anthropic's settlement with authors and other rightsholders who accused the company of using pirated books to build Claude. Reuters reported that the deal resolves Bartz v. Anthropic and is the largest known copyright settlement ever recorded in an American court. Authors and publishers covering more than 92% of the roughly 480,000 works in the case filed claims to collect.
You don't need a law degree to see what happened here. Anthropic won the big legal question and still had to write the biggest check in copyright history. That is the part every AI company should be reading twice. Now-retired Judge William Alsup ruled last June that training Claude on the authors' books was fair use, a finding AI companies from OpenAI to Meta have been watching closely ever since. But Alsup drew a hard line on how Anthropic got those books in the first place. The company had downloaded more than 7 million pirated titles from LibGen and PiLiMi into what it internally called a central library. Used or not, it didn't matter.
That split is the whole story. Fair use protected the training. Piracy did not.
What the Settlement Pays
Under the settlement, rights holders whose books qualify are set to receive roughly $3,000 per title, split among the authors and publishers who share rights to a given work. That is real money. It is also much less than many writers believe their books are worth after seeing them swept into AI systems without permission.
Mary Rasenberger, CEO of the Authors Guild, called the settlement an excellent result for authors, publishers and rightsholders generally, and said it sent a strong message to the AI industry about the consequences of pirating authors' work to train models. That statement checks out. It also captures the awkward bargain at the center of the case: the payout is historic in the aggregate, while the per-book number is still small enough to anger individual authors.
Reuters reported that plaintiffs' attorneys at Susman Godfrey had asked for $187.5 million in fees. Martinez-Olguin awarded them a little more than $101 million instead, trimming the request by nearly half. Some class members objected that the payout was too small. Others said the attorneys were taking too much, or that the deal wrongly excluded certain copyright owners. Martinez-Olguin overruled those objections, writing that complaints about the settlement's size were not grounded in a realistic assessment of the risks and rewards of a trial.
That's a judge telling authors something blunt. A jury trial against a well-funded defendant, after that defendant already had a fair use ruling in hand, was never going to be a sure thing.
The Rest of the Industry Is Watching
Anthropic's case is the first major AI book-piracy suit to actually close. Every other AI company facing one is reading the outcome closely. Authors including Sarah Silverman and Ta-Nehisi Coates have sued OpenAI and Meta over similar claims. Paul Tremblay's proposed class action against OpenAI is still pending, and so is the Authors Guild's own case against the company.
The pressure is not only coming from individual authors. Reuters reported this month that Hachette, Cengage and Elsevier - along with author Scott Turow - sued Google in New York, accusing it of using copyrighted books to train Gemini. Five major publishers also sued Meta in May over Llama training. The industry is no longer waiting around for one clean Supreme Court answer. It is filing case after case and forcing AI companies to defend the plumbing behind their models.
Six authors who opted out of the Anthropic deal filed their own lawsuit in December against Anthropic, OpenAI, Google, Meta, xAI and Perplexity. Bloomberg Law reported that the group includes John Carreyrou and seeks statutory damages of up to $150,000 per infringed work. Do the math on 480,000 books at that rate, and you'll understand why Anthropic's board chose a $1.5 billion settlement over a trial that could have become much worse.
Anthropic is backed by Amazon and Alphabet, two companies with their own reasons to want AI copyright law settled rather than fought book by book in court. OpenAI and Meta have deep-pocketed backers too, but neither has a settled price yet. Anthropic just set one. Frankly, it is high enough to make piracy look less like a clever shortcut and more like a liability line item that got out of hand.
Also read: Oracle Sues Wisconsin Over A $7 Billion Data Center Collateral Bill • A Judge Just Capped Sony's Copyright Case Against Udio at 333 Songs • BlackRock Sells $12 Billion in Bonds to Fund Meta's Texas AI Data Center
