Judge grants final approval to Anthropic's $1.5 billion author copyright settlement
US District Judge Araceli Martinez-Olguin signed off Monday on Anthropic's $1.5 billion settlement with authors whose books the company downloaded from pirate libraries to train Claude. Bloomberg Law called it the largest copyright class action settlement in US history and the first AI training copyright lawsuit to reach full resolution.
What the settlement covers
About 480,000 works fall within the class, with payouts running roughly $3,100 per title, per Bloomberg Law. Rights holders collect their share whether or not they submitted a claim, subject to claims administration. Judge Martinez-Olguin also cut class counsel's fee request from $187.5 million to $101.6 million, a reduction of roughly $86 million from what plaintiffs' attorneys sought, per Law.com.
What drove the case was narrower than the headline figure suggests. Earlier presiding judge William Alsup (now retired) had ruled that using copyrighted text to train an AI model constitutes fair use. Illegal, in Alsup's findings, was how Anthropic sourced some of those texts: titles downloaded from Library Genesis and the Pirate Library Mirror rather than purchased. That distribution rights violation was heading to trial when Anthropic agreed to settle.
Why it matters
Approval closes this individual case but creates no binding legal precedent for the rest of the industry. Because Anthropic settled before an appeal, Alsup's district-court fair-use ruling remains persuasive authority only, per TechCrunch. Judges in suits against OpenAI, Meta, Google, and Midjourney can each reach their own conclusions on their own facts.
For AI product teams pricing training data risk, roughly $3,100 per pirated title is now a concrete number on the table. Millions of books were involved in Anthropic's downloads, which shows the scale of exposure companies face from similar sourcing decisions. Publishers are pressing ahead elsewhere: a new class action against Google over Gemini training data was filed July 14 by Hachette, Cengage, Elsevier, and author Scott Turow, per TechCrunch.
What to watch next
OpenAI and Meta each face parallel author suits with no announced settlement. At $3,100 per work, both sides now have a reference point for negotiations. Also watch whether content licensing deals follow: settlement class members include major rights holders who could redirect from litigation to structured data licensing arrangements with AI labs.