JUDGE FINDS ANTHROPIC'S AI TRAINING ON BOOKS IS FAIR USE

A federal judge in California ruled that Anthropic, an Al company backed by Amazon and Alphabet, did not violate copyright law by using books to train its Claude large language model.

- 

The decision, penned by U.S. District Judge William Alsup, declared the training process “fair use” under U.S. copyright law, as it transforms copyrighted works into new, original outputs without reproducing the authors’ creative elements. The ruling is a significant win for Al companies facing a wave of lawsuits over how they use copyrighted material to build their models.

The case, brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, accused Anthropic of using their books without permission to train Claude, a chatbot designed to generate human-like text. The judge found that this use—feeding books into Al to learn patterns and produce unique responses—falls under fair use, likening it to a reader studying a book to become a better writer. The decision draws on legal precedents like the Google Books case, where scanning books for search purposes was deemed transformative.

However, the judge drew a hard line at Anthropic’s storage of over 7 million pirated books in a “central library,” ruling this act infringed copyrights. A trial is set to determine damages for the piracy, with potential penalties of up to $150,000 per work for willful infringement. Anthropic’s attempt to justify downloading books from pirate sites, arguing it later bought legal copies, didn't clear it of liability, though it may reduce damages.

THE FAIR USE ARGUMENT