A chapter that half the AI industry watched with bated breath is now closed. On Monday in San Francisco, US District Judge Araceli Martinez-Olguin granted final approval to Anthropic’s $1.5 billion settlement. It’s the largest known copyright settlement in US history — and the first major case where an AI company settles with authors over how it trained its models.
What it was about
A group of authors accused Anthropic of illegally obtaining their books and storing them in a central digital library to train Claude. The settlement covers nearly 500,000 copyrighted works, with a potential payout of around $3,000 per work. More than 91 percent of eligible authors and publishers have submitted claims.
Objections overruled
The deal wasn’t without pushback. Some authors thought the amount was too small, others argued the plaintiffs’ attorneys were overcompensated or that certain rights holders were wrongly excluded. Martinez-Olguin overruled all of those objections. She awarded the attorneys more than $101 million — of the $187.5 million they had requested.
Now-retired Judge William Alsup gave the deal preliminary approval last September. Now it’s final.
My take
$1.5 billion sounds like a lot — and it is. But measured against how central training data is to a frontier model, it reads more like the price of legal certainty than a penalty that actually stings. That’s exactly what makes the case matter: it puts a number on a problem the whole industry shares. Anyone training on other people’s work now has an order of magnitude that plaintiffs and defendants can anchor to. It’s the first big settlement of its kind — it definitely won’t be the last.
Sources: