Key holding or allegation
The court held the LLM training uses before it highly transformative and fair use, and treated digitisation of purchased books as fair use, while analysing pirated acquisition/retention separately and refusing fair use for that conduct on the summary-judgment record.
Why it matters
Landmark U.S. merits ruling separating transformative training from the legality of source acquisition and library retention.
Facts
Authors challenged Anthropic's acquisition, digitisation, retention and use of books to build its library and train LLMs.
Issues before the court
Whether training, digitisation of purchased books and acquisition/retention of pirated books were fair use.
What the court did not decide
Does not create a categorical rule that all AI training is fair use or that unlawful acquisition is cured by later transformative training.
Procedural timeline
U.S. District Court for the Northern District of California
Summary-judgment fair-use order: LLM training uses were held fair use; digitisation of purchased books for the central library was also fair use; pirated acquisition/retention was treated separately and did not receive fair-use protection on the record.
U.S. District Court for the Northern District of California
Final settlement approval: Court granted final approval to the class settlement; the earlier fair-use merits ruling remains part of the doctrinal record.