In a recent ruling, a federal judge in San Francisco declared that Anthropic’s utilization of books without permission to train its artificial intelligence system was deemed legal under U.S. copyright law. U.S. District Judge William Alsup sided with tech companies on this crucial matter for the AI industry, asserting that Anthropic’s use of books by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson to train its Claude large language model constituted “fair use.” However, the judge also noted that Anthropic’s copying and storage of over 7 million pirated books in a centralized library infringed upon the authors’ copyrights and was not fair use. Consequently, a trial has been scheduled for December to determine the amount Anthropic owes for this infringement, with U.S. copyright law permitting statutory damages of up to $150,000 per work for willful copyright infringement.
An Anthropic spokesperson expressed satisfaction with the court’s recognition of their AI training as “transformative” and aligned with the core principles of copyright aimed at nurturing creativity and scientific advancement. The lawsuit, initiated by the writers as a proposed class action against Anthropic, alleged that the company, supported by Amazon and Alphabet, unlawfully used pirated versions of their books to teach Claude how to interact with human prompts. This legal action is part of a broader trend where authors, media outlets, and copyright holders have taken legal action against companies like OpenAI, Microsoft, and Meta Platforms over their AI training practices.
The concept of fair use plays a pivotal role in allowing the use of copyrighted material under certain circumstances without the explicit permission of the copyright owner, serving as a key defense for tech companies in these disputes. Alsup’s ruling is the first to address fair use in the context of generative AI, where AI entities argue that their activities are transformative and beneficial for innovation, while copyright owners view them as infringing upon their rights and threatening their livelihoods.
Acknowledging the transformative nature of Anthropic’s AI training, Alsup emphasized that it aimed to create original content rather than replicate existing works. However, the judge also criticized Anthropic for maintaining pirated copies of books in a universal library, a practice that the authors contended was unauthorized and detrimental to their interests. Notably, Anthropic and other leading AI firms have faced accusations of downloading pirated copies of numerous books to train their systems, prompting legal scrutiny and debate around the boundaries of fair use and copyright compliance in the AI industry.
“Alsup stated on Monday that the precedent set by the ruling was crucial for any future fair use cases,” reported Blake Brittain in Washington. The article was edited by Alexia Garamfalvi, Chizu Nomiyama, Louise Heavens, and Matthew Lewis.
