A federal judge in California issued a landmark ruling on Monday in one of the first major court decisions addressing Artificial Intelligence (AI) training and copyright law. In a mixed ruling, the court found that training large language AI models (LLMs) on copyrighted books is legal under the doctrine of fair use, while also holding that AI platforms cannot use pirated materials to train their systems.
The lawsuit, filed by three authors, challenged the use of copyrighted materials by Anthropic, a leading AI company. Anthropic reported over $1 billion in annualized recurring revenue at the end of 2024, and is most known for its platform “Claude.” The authors contend that Anthropic used copyrighted books without permission to train Claude’s family of LLMs. The lawsuit shows that Anthropic has used several million books to train its systems — some were purchased in print form and then digitally scanned, while others were pirated from online sources.
Writing for the US District Court for the Northern District of California, Judge William Alsup ruled that converting legally purchased books to a digital format for the purpose of training LLMs does not infringe on copyright protections because it was merely a format change, and “was not done for purposes trenching upon the copyright owner’s rightful interests.” The ruling notes that Anthropic’s use of the copyrighted materials falls under the fair use doctrine because of its “transformative” nature. The Copyright Act of 1976 provides four factors that courts may use in determining if the use of copyrighted material is legal:
(1) The purpose and character of the use, (2) the nature of the copyrighted work, (3) the amount and substantiality of the portion used, and (4) the effect of the use on the potential market.
In 1994, the Supreme Court ruled that when copyrighted materials are used to create something new and transformative, the purpose and character of the use make it lawful. This legal framework is further supported by Article 1 of the Constitution, which grants Congress the power to create copyright laws that “promote the progress of science and useful arts.”
Monday’s ruling sets a key precedent and presents a major victory for AI companies. However, the ruling makes a distinction that only legally acquired source material can be used to train LLMs. The ruling grants summary judgment for Anthropic on its use of purchased books, but the counts of pirating material will proceed to a trial to determine damages.
This comes amidst a rapidly evolving landscape surrounding AI and copyright law. As these laws continue to evolve, broader concerns over AI and its intersection with the rule of law remain a topic of significant interest.
English (US) ·