How U.S. Copyright Law Applies to LLM Training

A U.S. court analyzes whether Anthropic’s use of copyrighted books to train Claude AI qualifies as fair use. Drawing from past cases like Google Books and Texaco, the court finds that training LLMs is likely “transformative,” favoring fair use—so long as outputs don’t infringe. Authors argue against the legality of even format changes, but the court emphasizes precedent and public benefit. The first fair use factor strongly supports Anthropic’s position, though other uses remain under review.

ANALYSIS

Section 107 of the Copyright Act identifies four factors for determining whether a given use of a copyrighted work is a fair use:

[T]he fair use of a copyrighted work . . . for purposes such as criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copyright. In determining whether the use made of a ...