On May 22, 2025, in Bartz v. Anthropic PBC, a federal judge in the District Court for the Northern District of California, William Alsup, was inclined to find that Anthropic had infringed copyright law in making the initial copies of the pirated book, but that its subsequent use in training its generative AI models was fair use (Case No. 24-cv-05417).

Anthropic argued that the four factors of the fair use analysis - purpose of use, nature of the copyrighted work, reasonableness of the use, and market impact - all weighed in its favor, claiming that its use was “extremely transformative. “. The authors countered that while the courts have ruled on fair use many times, the act of downloading a copyrighted work from a pirate site to avoid payment of royalties does not justify fair use.

Judge Alsup said the court might find that Anthropic should pay for the initial copies it obtained, suggesting that Amazon book prices could be used to determine costs. However, he pointed to the relatively low cost of legal alternatives, stating that he had found books at library sales for $1. Judge Alsup also challenged the author's argument about the impact of AI on the book market, noting that Anthropic had erected guardrails to prevent verbatim output.

As a result, Judge Alsup asked the parties to submit supplemental briefs on a 2015 Second Circuit decision on which Company A relies heavily, Authors Guild v. Google Inc. and whether that decision held that Google's original copying in that case was protected by fair use protection, filed a supplemental brief.
The judge at the summary judgment hearing three weeks ago in which the writers accused Meta of copyright infringement said Meta's AI model Llama could destroy the market for copyrighted works, but no ruling has been made because of insufficient evidence from the writers about that potential impact.