rotative
the decoder · July 22, 2026

Anthropic pays $1.5B for pirated books, but AI training wins fair use

anthropic settled a class action for $1.5 billion for using pirated books, but the court ruled ai training on legally obtained data is fair use.

read in 简体中文 · 繁體中文 · 日本語 · 한국어

a federal court approved a $1.5 billion settlement against anthropic for downloading nearly half a million pirated books, requiring the company to destroy the files and pay about $3,000 per claim. despite the record penalty, the ruling distinguishes between piracy and training, affirming that ai training on legally obtained materials is fair use. this creates a precedent that shields ai labs from liability for data scraping, provided the initial acquisition was legal. the decision leaves unresolved questions regarding whether mass web scraping constitutes legal acquisition, suggesting future litigation will focus on consent rather than training itself.

read at source ↗

← rotative