Judge rules Anthropic’s AI training with books is fair use but storing pirated copies violates copyright

In a significant ruling from a federal court in San Francisco, Anthropic, an AI company, was granted a victory regarding its training practices for the large language model, Claude. The court determined that the use of copyrighted literature for AI training falls under the fair use doctrine as per U.S. law. However, the situation is complicated by Anthropic's retention of millions of pirated books, which the court asserted constitutes a violation of copyright laws. U.S. District Judge William Alsup emphasized that the incorporation of works by authors such as Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson into the training of the AI was "exceedingly transformative." He argued that this use aligns with the overarching goals of copyright law, which is to promote creativity and innovation. He stated, "Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works not to race ahead and replicate or supplant them but to turn a hard corner and create something different." However, the judge did not hold back in criticizing Anthropic’s storage of over 7 million pirated books, referring to it as a “central library” that does not qualify for fair use protections. This aspect of the case highlights a stark contrast between the transformative nature of AI training and the unlawful retention of copyrighted materials. The authors involved in the case, who filed a proposed class action last year, argued that Anthropic had utilized pirated versions of their works without obtaining permission or providing compensation. With backing from major tech firms like Amazon and Alphabet, Anthropic expressed approval of the court’s findings regarding their AI training. A spokesperson stated, “We are pleased that the court recognised our AI training as transformative and consistent with copyright’s purpose in enabling creativity and fostering scientific progress.” This ruling marks a pivotal moment in U.S. legal discourse surrounding fair use in the realm of generative AI, a defense strategy increasingly employed by companies such as OpenAI, Meta, and Microsoft amidst ongoing copyright disputes. While AI developers argue that utilizing copyrighted material fosters innovation through transformative works, copyright holders assert that such practices jeopardize their financial interests. As the case progresses toward a December trial, Anthropic could face statutory damages reaching up to $150,000 per work for willful infringement. The outcome of this trial is poised to shape future legal frameworks regarding how AI enterprises access and employ protected creative works in developing generative models.