A federal judge has ruled in favor of AI company Anthropic, declaring that training its AI models solely on purchased books without authors’ permission is fair use. However, the decision only applies to physical books purchased by Anthropic and does not address whether the outputs of an AI model infringe copyrights.
The lawsuit was filed by writers Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who claimed that Anthropic trained its AI models on pirated material. Judge William Alsup determined that Anthropic’s actions were fair use, citing the Copyright Act as the reason for this ruling. He also stated that using printed books to train an AI model was transformative and did not infringe on authors’ original works.
However, the judge ruled that a separate trial is necessary to determine damages for millions of pirated book copies stored in Anthropic’s central library. The court emphasized that downloading source copies from pirate sites without purchasing them lawfully does not meet the requirements for fair use.
Anthropic CEO Jennifer Martinez released a statement expressing satisfaction with the ruling, emphasizing that its AI models are used to create new and original works rather than replicate existing ones.
Source: https://www.theverge.com/news/692015/anthropic-wins-a-major-fair-use-victory-for-ai-but-its-still-in-trouble-for-stealing-books