Anthropic Settles Copyright Lawsuit for $1.5B

The US Northern District Court of California formally approved the final settlement of the Bartz v. Anthropic copyright class action. Anthropic will pay $1.5 billion to the authors, while the court awarded about $101.56 million in attorneys' fees. This event has drawn widespread attention not only for setting the highest known payout in US copyright history but also for providing a highly valuable legal precedent for AI model training copyright disputes.

Key Details and Legal Boundaries

The case originated from authors, including novelist Andrea Bartz, suing Anthropic for using over 7 million copyrighted books to train its AI assistant, Claude. According to @koltregaskes, internal emails showed Anthropic knowingly downloaded data from pirate libraries like Library Genesis and Pirate Library Mirror. The presiding judge clarified a crucial legal boundary: using these books for model training itself qualifies as "fair use," but Anthropic's act of acquiring and storing these pirated books from illegal sources remains a violation, which is the fundamental reason for the massive settlement. The settlement roughly covers 500,000 books, averaging about $3,000 per book, and requires Anthropic to destroy its stored pirated copies.

Reactions and Cost Comparisons

This astronomical penalty sparked intense discussion within the AI community. User @eliebakouch compared it to the training cost of DeepSeek V3, pointing out that the settlement is 269 times higher, highlighting the immense financial risks top AI companies face regarding data acquisition and compliance. Media outlets like Reuters and industry observers noted that the legal risks surrounding training data have evolved from mere disputes into substantial business costs capable of shaking the industry. Furthermore, some authors and publishers opted out of the settlement to continue their lawsuits against Anthropic.

2026-07-21 ~ 2026-07-23 · 23 related posts

Primary sources

14 near-duplicate retellings: Apprehensive_Sky1950 · rohanpaul_ai · davidavvv · gaganghotra_ · CodeByPoonam · koltregaskes · LuizaJarovsky · LuizaJarovsky · Distinct-Question-16 · luisdans · zijing_wu · ctjlewis · ArtificialOther · secemp9