Anthropic agrees a $1.5 billion copyright settlement

Anthropic agreed to pay $1.5 billion to settle Bartz v. Anthropic, a class action brought by book authors over the company’s use of pirated texts to train its Claude models, a figure reported as the largest copyright recovery in US history. Under the proposed settlement, authors covered by the class stood to receive roughly $3,000 for each of the approximately 500,000 works involved, well above the $750 statutory minimum for proven copyright infringement, and Anthropic agreed to destroy the pirated files it had downloaded.

The settlement followed a June 2025 ruling by Judge William Alsup of the Northern District of California, who had found on summary judgment that training Anthropic’s models on books it had lawfully purchased was fair use, while denying the company summary judgment on its separate practice of downloading books from pirate sites such as Library Genesis and keeping them in a permanent digital library. That narrower question — Anthropic’s liability for the piracy itself, regardless of whether the pirated copies were later used in training — had been heading to trial, where statutory damages could in theory have run into the hundreds of billions of dollars; the settlement resolved it instead.

The deal was announced as an agreement in principle, subject to court approval of the settlement class and the distribution plan, a process that included identifying and notifying authors of roughly 500,000 works and took most of the following year; a federal judge granted final approval in July 2026, after trimming the requested attorneys’ fees. For other AI companies facing similar suits over training data — including OpenAI, Meta and Stability AI — the settlement supplied a concrete benchmark, a per-work dollar figure, that plaintiffs’ lawyers elsewhere could point to in their own negotiations, even though the underlying legal question of whether training itself infringes copyright remained unresolved outside this one case.