According to 404 Media, Amazon is cutting off the spines of rare books and scanning them for AI training data. Reporters placed a tracking device in a rare book that ultimately arrived at an Amazon facility in Las Vegas called VGT3, identifiable by a symbol of a dinosaur holding a book. Amazon told 404 Media it "purchases books through commercial channels to improve the products and services customers use."
AI companies require massive amounts of text to train their large language models, which have already consumed much of the available internet content. Rare books, especially those out of print or unavailable online, represent a valuable new source of training data since content published before 2022 was not generated by AI.
This practice has grown as companies seek to avoid "model collapse," a phenomenon where AI output quality degrades after training on too much AI-generated text. Anthropic has also reportedly used illegally pirated books for AI training purposes.