A federal district court just ruled that training #AI tools on copyrighted texts is fair use.
https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/
#Anthropic is in the clear for training #Claude on millions of books without permission or payment. However, it's liable for obtaining many of those books illegally, and its settlement of those claims (about $3k/title) was approved by the court.
So far this decision only applies in the Northern District of California. But it breaks the ice and will (as lawyers say) be "persuasive but not binding" in other districts.