Anthropic's $1.5B author copyright settlement wins final approval
TL;DR
- At $1.5 billion across roughly 500,000 works, this is the largest known copyright recovery in U.S. history, per the authors' lead counsel.
- Judge Alsup ruled training on copyrighted text is fair use, but acquiring those books from pirate sites (Library Genesis, Pirate Library Mirror) is not.
- Because Anthropic settled before appeal, Alsup's fair-use ruling binds no future court, leaving OpenAI, Meta, and Google to litigate that question independently.
A federal judge in San Francisco has signed off on the biggest copyright cheque the tech industry has ever written. On July 20, U.S. District Judge Araceli Martinez-Olguin granted final approval to Anthropic's $1.5 billion settlement with a class of book authors, Bloomberg Law reported, closing out the first major U.S. AI-training copyright case to actually resolve.
The distinction that made the number possible is worth pausing on. Now-retired Judge William Alsup ruled last June that training Claude on authors' books was fair use. What was not fair use, in his reading, was Anthropic's decision to save more than 7 million pirated books to a 'central library' that would not necessarily be used for AI training. The settlement covers more than 480,000 works and pays roughly $3,100 per title, according to Yahoo Finance, with authors and rights holders filing claims covering about 92% of eligible titles.
Why this matters if you are not a copyright lawyer: for two years the industry has argued that scraping the open internet is basically a data problem, not a legal one. Alsup's split ruling, and now the price tag attached to it, reframe it as a supply-chain problem. Fair use may still cover the training step itself, but where you got the corpus is now a discoverable, dollar-denominated liability. Lead plaintiffs' counsel Justin Nelson called it 'the largest known copyright recovery in history,' per WSAU's report.
The honest caveat is that this is a settlement, not a precedent. Nothing here binds the judges hearing the parallel suits against OpenAI, Meta or Google, and some authors have opted out to pursue their own claims. What the reporting also does not tell us is how Anthropic plans to fund the tranches on its current cost base, or whether other content classes such as news, code or images will follow the same per-work math.
The upside, if you are on the rights-holder side of the table, is that a floor has been set. If you are running an AI lab, the takeaway is quieter: dataset hygiene just went from a legal-team footnote to a line item the CFO has to model.
What others are reporting
-
TechCrunch Read →
Explains why Alsup's fair-use ruling will not bind other courts: Anthropic settled rather than appealing, and liability rested on piracy acquisition, not training.
-
SiliconANGLE Read →
Names the pirate sources (Library Genesis, Pirate Library Mirror) and carries the presiding judge's quote on why the settlement amount is justified.
The $1.5 billion settlement provides substantial benefits to the class in light of the novel claims asserted.
-
Authors Guild Read →
First-party plaintiff-side guide covering copyright registration requirements, the default 50/50 author-publisher split, and the four-installment payment schedule through 2027.
Each title will be paid an equal amount, with approximately 500,000 eligible titles sharing the $1.5 billion settlement fund.
-
Reuters Read →
Wire report with the authors' lead attorney on record; frames the ruling as the largest known copyright recovery in U.S. history.
It is the largest known copyright recovery in history.
Originally reported by news.bloomberglaw.com
Read the original article →Original headline: US Judge Grants Final Approval to Anthropic's $1.5B Copyright Settlement With Book Authors