Home / Technology / AI Firm Accused of Pirating Millions of Books
AI Firm Accused of Pirating Millions of Books
31 Aug
Summary
- Music publishers allege Anthropic pirated over 7 million books for AI training.
- The $1.5 billion settlement is deemed insufficient by publishers.
- Allegations include torrenting lyrics and sheet music, harming songwriters.

Leading music publishers, including Sony, EMI, and Warner Chappell, are contesting Anthropic's $1.5 billion settlement, deeming it insufficient compensation for the alleged pirating of over 7 million books. The publishers claim Anthropic engaged in a massive campaign of illegal torrenting, scraping, and downloading copyrighted works, including lyrics and sheet music, to train its AI models. This alleged piracy began as early as July 2021, with co-founder Benjamin Mann reportedly using BitTorrent to download books from Library Genesis. Despite legal challenges to pirate libraries like Z-Library, publishers assert Anthropic accessed the content through "Pirate Library Mirror" (PiLiMi), pointing to internal communications as evidence of intentional reliance on piracy.
While Anthropic denies using pirated books for its commercial Claude AI models, publishers suspect a connection through "pretraining" phases involving synthetic data. They highlight that Anthropic continued using the LibGen dataset for quality control even after ceasing direct training on it. Publishers argue that Anthropic's actions, which include allegedly destroying physical books to create unauthorized digital copies of songbooks and sheet music, directly harm songwriters. They believe AI-generated songs competing on music charts, trained on uncompensated works, dilute royalty pools and threaten artists' livelihoods. Anthropic maintains that training generative AI is a "transformative fair use" and plans a robust defense.