Judge approves $1.5B Anthropic settlement for pirated books used to train Claude

Settlement terms and payout structure

The settlement provides $1.5 billion total, with roughly $3,000 per eligible title, and class counsel fees were cut in half. According to commenter @ilamont, the payout per eligible title is $3,000; for a traditional publishing contract involving one author the amount will be split down the middle. The judge slashed the class counsel’s fee from 12.5% ($187.5 million) to 6.8% ($101 million). The class counsel’s unreimbursed litigation expenses were $2.6 million. The three class representatives receive $15,000 each. Commenter @chrisaiv adds that the $1.5 billion is paid in installments and counsel’s fees are paid in step with those installments, making the class effectively Anthropic’s creditor with a direct financial interest in the company staying solvent through the payment schedule.

Judicial background on liability and fair use

Judge Alsup previously found Anthropic liable for piracy but held that training LLMs on the books constituted fair use. Commenter @driverdan points to Judge Alsup’s original order that determined liability for piracy while concluding that training LLMs on the books was fair use, and provides a link to the docket.

Reactions and critiques from commenters

Many commenters view the settlement as insufficient, calling it a "slap on the wrist" and comparing it to penalties for individual piracy. @BloodyIron wrote: "Did someone forget to consult with the MPAA and the RIAA on this one? This is a joke of an outcome. $3k per book. How much was it per song for Napster?" @pier25 noted that under US federal law pirating a single copyrighted work for commercial advantage carries up to five years in prison and a $250,000 fine, with higher penalties for repeat offenses or larger scale infringement. @processunknown described the outcome as "squarely in the 'cost of doing business' category."@gabriel666smith, identifying as an author, argued that the settlement is a tokenistic payment that does not establish ongoing royalties and compared it to the UK Public Lending Right system.

Implications for precedent and future AI training

The settlement does not create legal precedent, and some argue it establishes a financial moat for Anthropic while leaving authors with only a one‑time payment. @petterroea observed that a settlement means the case does not become legal precedent, calling it the best possible outcome for Anthropic given the scale of AI investment. @pjio said the settlement is insufficient for human authors but ideal for Anthropic, which now has a financial moat for others to train AI on those works (except for Chinese companies, which do not care either way). @sashank_1509 argued that a one‑time payment does nothing and that a royalty based on model output would be the correct legislative approach.

Global scope and limitations

The settlement applies only to U.S. rights holders, leaving authors outside the United States without compensation. @HackerThemAll warned that the case is Americentric, noting that the settlement resolves a U.S. case and assumes it is "cool for the entire world.""}