Anthropic’s $1.5B book-piracy settlement wins final approval — ~$3,000 per work

The largest copyright class-action settlement in US history just got a gavel. On July 20, Judge Araceli Martínez-Olguín granted final approval to Anthropic’s $1.5 billion settlement with authors and publishers over books downloaded from pirate libraries LibGen and PiLiMi to train Claude. Expected payout: roughly $3,000 per work before fees — about four times the common $750 statutory minimum for ordinary infringement.

What the ruling actually preserved

Judge William Alsup’s earlier mixed decision is the industry hinge: training AI on lawfully acquired copyrighted books can be fair use; building a library from pirate sites is not. Anthropic settled the piracy half to avoid a damages trial. Deputy GC Aparna Sridhar stressed that more than 91% of covered authors/publishers have already claimed — and that the fair-use half remains a landmark for the lab. Class counsel fees were cut from a requested $187.5M to about $101.6M (~6.8% of the fund).

Why every other lab is reading the footnotes

Dozens of AI copyright suits are still alive. This deal closes Bartz v. Anthropic without an appellate stress-test of Alsup’s fair-use holding — which is exactly why model trainers are smiling and many authors are not. Paying for piracy while keeping the training theory is the template OpenAI, Google, and Meta lawyers are already war-gaming.

Report: TechCrunch — Anthropic’s $1.5B settlement approved.

Fair use got a doctrine. Pirate torrents got an invoice. The rest of the docket is still loading.