The Piracy Problem: Where Fair Use Meets Theft
However, Anthropic’s legal victory comes with a significant caveat. While Judge Alsup ruled that the training itself constituted fair use, he ordered a separate trial to address allegations that Anthropic obtained many of these books through illegal means.
According to the lawsuit, Anthropic embarked on an ambitious project to create what it internally called a “central library” containing “all the books in the world” for permanent storage. The plaintiffs allege that millions of these copyrighted works were downloaded from pirate websites—an acquisition method that remains unambiguously illegal regardless of how the materials are subsequently used.
“We will have a trial on the pirated copies used to create Anthropic’s central library and the resulting damages,”Judge Alsup wrote in his decision. “That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for theft but it may affect the extent of statutory damages.”
This distinction creates a nuanced legal landscape where the use of copyrighted materials for AI training might be permissible, but obtaining those materials through illegal channels remains actionable. The ruling suggests that AI companies must be more careful about their data acquisition methods, even if their ultimate use of that data receives fair use protection.
