Science USA

U.S. Judge Approves AI Training on Copyrighted Works, But Requires Trial Over Pirated Books

U.S. Judge Approves AI Training on Copyrighted Works, But Requires Trial Over Pirated Books
Source: AP Photo
  • PublishedJune 29, 2025

 

A U.S. federal judge issued a landmark ruling in favor of AI company Anthropic, declaring that the use of copyrighted literary works to train its Claude language model constitutes “fair use.” However, the same ruling mandates a December trial to determine whether Anthropic illegally downloaded pirated books to create its training library.

Judge William Alsup of the Northern District of California found that Anthropic’s use of books it lawfully purchased and digitized was “exceedingly transformative.” In his opinion, the process mirrors that of an aspiring writer who studies existing works not to replicate them, but to inform new, original creations—deeming it permissible under fair use doctrine.

Yet, the court ruled differently on the issue of Anthropic’s initial acquisition of material. The company admitted to downloading over seven million books from piracy websites before transitioning to legally purchased copies. Alsup determined this centralized storage of pirated content was not protected by fair use and will be subject to a jury trial in December to assess related liability and damages.

Anthropic defended its practices, stating the AI training process aligns with copyright goals that encourage creative and scientific advancement. The company also signaled confidence in its legal position regarding both fair use and acquisition practices.

The ruling represents a major legal milestone for the AI industry. It is the first federal decision to affirm that training large language models using copyrighted text can fall under fair use—so long as the material is not pirated. Tech firms like OpenAI, Meta, and Microsoft, currently facing similar lawsuits, are watching closely.

While the ruling supports AI development, it highlights a clear divide: lawful acquisition through purchase and digitization can be permissible, but using unauthorized, pirated content remains problematic. The upcoming December trial will be critical in shaping financial and operational implications for companies using AI training data.

With input from Al Jazeera

Michelle Larsen

Michelle Larsen is a 23-year-old journalist and editor for Wyoming Star. Michelle has covered a variety of topics on both local (crime, politics, environment, sports in the USA) and global issues (USA around the globe; Middle East tensions, European security and politics, Ukraine war, conflicts in Africa, etc.), shaping the narrative and ensuring the quality of published content on Wyoming Star, providing the readership with essential information to shape their opinion on what is happening. Michelle has also interviewed political experts on the matters unfolding on the US political landscape and those around the world to provide the readership with better understanding of these complex processes.