What happened
On July 20, 2026, US District Judge Araceli Martínez-Olguín (San Francisco) granted final approval to the $1.5 billion class-action settlement between Anthropic and a class of book authors/publishers over the company's use of pirated books (from Library Genesis and Pirate Library Mirror) to train Claude. This finalizes what was preliminarily approved by Judge Alsup in September 2025, following a May 2026 fairness hearing. Over 90% of the ~480,000 eligible works were claimed. Anthropic must pay ~$3,000/work in four installments through September 2027 and destroy the pirated source files; the settlement covers only past acquisition of training data, not model outputs or future training licenses.
Why it matters
This is the largest copyright recovery in US history and the first major US AI-training copyright case to be resolved with a payout rather than trial, setting a market price/template for the dozens of pending suits against OpenAI, Meta, Microsoft and Google. It operationalizes the prior fair-use/piracy distinction (training can be fair use; sourcing from pirate libraries is not) as a concrete financial liability model that will shape data-provenance practices industry-wide.
Action needed
AI developers should audit training-data provenance, avoid pirate/shadow-library sourcing, and evaluate licensing costs against statutory-damage exposure (up to $150k/work for willful infringement) revealed by this settlement's scale.