AI burns books? Tech giants are buying, shredding books to train smarter chatbots

AI companies are buying and destroying rare books to train models, aiming to prevent AI-generated content contamination. Reports from 404 Media and the Washington Post detail how firms like Anthropic buy books, scan them, and shred originals. A class action lawsuit against Anthropic resulted in a $1.5 billion settlement. A judge ruled that destroying physical copies made digital copies replacements, constituting fair use. Critics, including former officials and creators, condemn the practice. ISBNdb denies involvement in purchasing books for AI training.

AI companies are buying and destroying millions of rare books to train their models, according to a report by 404 Media. The books are scanned and then shredded to prevent AI-generated content from contaminating training data. This practice aims to avoid 'model collapse,' where AI models degrade when trained on AI-generated output. The Washington Post reported in January that Anthropic attempted to buy millions of books, slice their spines, and scan pages for its chatbot Claude. This led to a class action lawsuit by authors, which Anthropic settled for $1.5 billion. Judge William Alsup ruled that destroying the physical books made the digital copies replacements, thus constituting fair use. Critics, including former US representative Brad Carson and creator-rights advocate Ed Newton-Rex, condemn the destruction, arguing it rewards destruction and penalizes preservation. Elon Musk has asked SpaceX AI team to preserve rare books. ISBNdb, a company that helps with book purchases, disputes reports of its involvement in AI training purchases.