On June 24, 2025, Senior District Judge William Alsup ruled in favor of Anthropic in a pivotal AI copyright case, determining that training its AI models on legally purchased books without the authors' permission constitutes fair use.

The court concluded that training large language models on copyrighted works was 'exceedingly transformative' under Section 107 of the Copyright Act.

This class action lawsuit was initiated by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who alleged that Anthropic used pirated versions of their works without permission or compensation.

Anthropic reportedly downloaded over seven million pirated books and scanned pages from legally purchased copyrighted books for its AI training.

Judge Alsup highlighted that Anthropic's executives knowingly accessed unauthorized copies to avoid legal and business complications, preferring pirated books over legally purchased ones.

Fair use interpretations are complex and often hinge on factors such as the purpose of use, commercial gain, and the transformative nature of derivative works.

The lawsuit did not find evidence that Anthropic's AI directly reproduced specific content from the authors' books, as it filtered user inputs and outputs to avoid direct infringement.

Potential damages from the piracy claims could reach billions, given the minimum statutory damage of $750 per infringed book.

The case remains ongoing, with the authors potentially appealing the unfavorable summary judgment after the trial concludes.

During a hearing, Judge Alsup suggested a settlement, emphasizing that purchasing a book after initially pirating it does not absolve Anthropic of liability.

AI companies argue that their use of copyrighted materials transforms it into new content, while copyright owners assert that such practices threaten their livelihoods.

This ruling contrasts with previous cases, such as Thomson Reuters v. Ross, which found against fair use, highlighting the evolving landscape of AI copyright law.