Legal risk / Jul 21, 2026 / 4 min
The Fair-Use Win Nobody Can Cite
On July 20, a federal judge gave final approval to Anthropic's $1.5 billion copyright settlement — the largest in U.S. history — after William Alsup ruled training AI on books is fair use, then Anthropic paid to avoid a jury trial on seven million pirated downloads that will never reach appeal.
Anthropic just closed the most important copyright case in AI history with a $1.5 billion check — and the legal win that matters most will never bind a court.
What's new: On July 20, U.S. District Judge Araceli Martinez-Olguin granted final approval to Anthropic's settlement in Bartz v. Anthropic — roughly $3,000 per work across an estimated 500,000 titles, the largest copyright recovery in American history. More than 91% of eligible authors and publishers have already claimed their share.
The split ruling that started it: Retired Judge William Alsup handed down the industry's most consequential fair-use decision in June 2025.
- Training on copyrighted books: Fair use. Alsup called Anthropic's use "exceedingly transformative" — the first major ruling that LLM training on published works can be legal without a license.
- Torrenting seven million books into a "central library": Not fair use. Alsup found Anthropic "had no entitlement to use pirated copies" and ordered a December trial on damages that could have reached hundreds of billions.
Anthropic settled the piracy count for $1.5 billion instead of rolling the dice with a jury.
Why the check is the tell: Statutory damages run up to $150,000 per work. At 500,000 titles, a loss could have bankrupted the company. The $3,000-per-book payout is four times the statutory minimum — generous for a class action, pocket change compared to what a jury might have awarded on willful piracy.
Martinez-Olguin overruled authors who called the deal too small, writing objections were "not grounded in a realistic assessment of the overall risks and rewards of a trial." She also slashed class counsel's fee request from $187.5 million to $101 million.
What Anthropic says: Deputy general counsel Aparna Sridhar: "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."
What plaintiffs' counsel says: Lead attorney Justin Nelson: "It is the largest known copyright recovery in history. We look forward to making distributions to the Class as promptly as possible."
Why nobody else can rely on this: Alsup's fair-use ruling lives in one Northern District of California file. Because Anthropic settled before trial and never appealed, no circuit court will review it. Other judges — starting with Loretta Preska in New York, where Hachette, Elsevier, and Scott Turow sued Google over Gemini training on July 10 — are free to disagree.
Who's still fighting: Authors who opted out before the February deadline — including the 100-plus plaintiffs in Shakespeare et al. v. Anthropic — are pursuing separate suits seeking up to $150,000 per registered work. Google, Meta, OpenAI, and Midjourney face their own training-data cases. The Bartz settlement covers past piracy only; it does not license future training or shield Claude's outputs.
Convina's view: Anthropic bought the cleanest possible IPO narrative — a headline fair-use win and a closed checkbook — while ensuring the precedent that would actually govern the industry died in settlement. Every rival lab will cite Alsup in pitch decks and ignore him in court until a circuit affirms or rejects training without permission. The $1.5 billion wasn't a penalty. It was the price of keeping fair use permanently provisional.