AI Firms Suspected of Buying Bulk Used Books for Data

Used book retailers in the UK, Ireland, and Australia report a surge in mysterious bulk orders for thousands of titles ^[1]. Booksellers suspect artificial intelligence firms are acquiring these physical volumes for "clean" training data ^[1][2]. This data is reportedly unpolluted by newer content generated by AI models ^[1].

Reports suggest companies like Anthropic may be scanning and then destroying these books ^[1]. These large-scale purchases cover diverse subjects to improve model reasoning and accuracy ^[3]. Retailers expressed concern over the potential loss of rare physical copies to corporate data processing ^[1][3].

The trend highlights the growing difficulty of finding authentic, human-generated text for future AI development ^[2]. While sales have increased, many independent shops worry about the long-term impact on literary preservation ^[1].