← Back to feed

A federal judge in San Francisco signed off on Anthropic’s $1.5 billion settlement with book authors on Monday, closing the first major AI copyright case in the United States and establishing the sharpest legal boundary yet between what AI labs can and cannot do with written works.

US District Judge Araceli Martinez-Olguin granted final approval of the settlement — the largest known copyright recovery in US history — rejecting objections that the sum was too small. The class covers more than 5,000 authors and publishers.

What The Courts Decided

The case split into two rulings on two separate questions.

In 2025, then-Senior Judge William Alsup ruled that Anthropic’s use of book text to train Claude constitutes fair use under copyright law. That ruling stands. AI companies can train on copyrighted books without licensing them.

The second question — what Anthropic did with the books beyond training — is where the company lost. Internal discovery revealed that Anthropic had assembled a “central library” of more than 7 million pirated books, most acquired via scraping of known piracy sources. Storing that library independently of training constituted copyright infringement. Potential damages at trial, had the case proceeded, were estimated in the hundreds of billions of dollars.

That exposure produced the settlement.

Payment Structure

The $1.5 billion is paid in tranches:

• $300 million paid after preliminary approval in September 2025
• $300 million due within five days of Monday’s final approval
• $450 million on the first anniversary of preliminary approval
• $450 million on the second anniversary

More than 91% of covered authors had already claimed their share by the time Monday’s ruling came down.

Judge Martinez-Olguin awarded $101 million in attorney fees — substantially less than the $187.5 million plaintiffs’ lawyers sought. A prior version of the settlement had been rejected by Judge Alsup specifically over the fee structure.

What Stays Open

The settlement resolves claims for the class. A group of authors and publishers opted out and have filed separate lawsuits against Anthropic that remain active. Monday’s ruling has no direct bearing on those proceedings.

The fair use precedent also leaves the field open for follow-on litigation. Anthropic’s statement from deputy general counsel Aparna Sridhar underscored that: “We reached this settlement in 2025, after the court’s landmark ruling that training AI on books is fair use under copyright law — which remains the law today.”

The Strategic Read

For AI labs, the outcome is nearly optimal. The training-is-fair-use ruling survives. The $1.5 billion, while large as a headline, is manageable for a company at a $965 billion valuation with a $30 billion annual revenue run rate. And the case clarifies the one genuinely risky behaviour: operating large pirated content repositories as infrastructure separate from any training pipeline.

Every major lab runs some form of data operation. The ruling creates a clear audit question — does the data get used for training, or does it sit in a library indefinitely — that legal teams can now answer cleanly.

The fair use question, the more consequential one for the industry, is settled for now. The authors who pushed for trial wanted that ruling revisited. Instead, it hardened.