On August 12, 2025, Judge Alsup of the Northern District of California denied Anthropic's request to certify for early appellate review the court's June 23, 2025, summary judgment ruling in Bartz et al. v. Anthropic PBC, 2025 U.S. Dist. LEXIS 154870 (N.D. Cal. Aug. 12, 2025).
In the June 23 order, the Court held that training an AI model on lawfully acquired and digitized books was transformative and thus a protected fair use, emphasizing that the purpose of the training is to teach the model rather than to serve as a substitute for the books themselves. At the same time, the Court drew a clear boundary: Anthropic's ingestion of pirated works from shadow libraries was not fair use. Judge Alsup noted that the wholesale downloading and permanent storage of millions of unauthorized works exceeded permissible use and posed serious risks of market harm. The June 23 ruling therefore split the claims—finding fair use for training on lawfully obtained materials, while sending to trial the claims related to copying and maintaining the pirated library.
Anthropic sought immediate appellate review of whether its use and storage of pirated books could be considered fair use, but the Court refused. Judge Alsup emphasized that the Ninth Circuit should not weigh in until a full factual record has been developed at trial. He also criticized Anthropic for "refus[ing] to come clean" about its acquisition and use of the pirated works, finding that those facts should be presented to, and weighed by, a jury.
Trial is currently set for December 1, 2025. The key issues will include whether the use and storage of pirated books can qualify as fair use and how to measure potential damages in the AI context.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
