A federal judge sides with Anthropic in lawsuit over training AI on books without authors’ permission
A significant legal decision has emerged from a federal court in a landmark case concerning artificial intelligence and copyright law. Federal judge William Alsup has ruled that it is legal for AI company Anthropic to train its large language models (LLMs) using published books without obtaining prior permission from the authors. This ruling marks a pivotal moment, as it is the first instance where a court has acknowledged the claim by AI companies that the “fair use” doctrine can absolve them of liability when copyrighted materials are used for training purposes.
This decision represents a substantial setback for authors, artists, and publishers who have initiated numerous lawsuits against prominent AI developers, including OpenAI, Meta, Midjourney, and Google. While this ruling does not automatically guarantee that other judges will follow Judge Alsup’s lead in similar cases, it establishes a foundational precedent that could significantly favor technology companies over creative professionals in future copyright disputes.
The core of these ongoing lawsuits often revolves around the interpretation of the fair use doctrine. This facet of copyright law is notoriously complex and has not been updated since 1976, predating the advent of the internet and, certainly, the concept of generative AI training datasets. Fair use assessments typically consider several factors: the purpose and character of the use (e.g., for parody or education), whether the material is reproduced for commercial gain, and how transformative the derivative work is compared to the original.
Companies like Meta have previously invoked similar fair use arguments in their defense against allegations of training on copyrighted works. However, before this week’s ruling, the judicial stance on such arguments remained largely uncertain.
In the specific case, Bartz v. Anthropic PBC, the plaintiff authors also raised questions regarding the manner in which Anthropic acquired and stored their works. The lawsuit alleged that Anthropic aimed to build a “central library” comprising “all the books in the world” for perpetual retention. Crucially, millions of these copyrighted books were reportedly downloaded without authorization from pirate sites, an act that is unequivocally illegal.
Despite granting that Anthropic’s use of these materials for training constituted fair use, the court has mandated a separate trial concerning the nature and acquisition of the “central library.” Judge Alsup explicitly stated in the decision, “We will have a trial on the pirated copies used to create Anthropic’s central library and the resulting damages.” He further clarified that while Anthropic might have later purchased legal copies of books initially stolen from the internet, this act would not negate liability for the initial theft, though it could influence the extent of statutory damages.
