Bulk book purchases in Japan linked to AI training operations
Japanese used bookstores are experiencing a surge in bulk sales, with some orders reaching 50 tons, reportedly destined for AI companies that scan and then destroy the books. Investigators suspect foreign AI firms are buying large quantities of Japanese literature for training data. The practice has raised concerns about the removal of important works from circulation.
The surge in bulk orders has been traced to a single logistics hub in Okayama Prefecture, though its operator declined to comment. One documented shipment weighed 50 tons and was sent to the U.S., where it reportedly entered Anthropic's scanning pipeline. Bookstore owners note that while novels and comics once dominated sales, the new bulk buyers focus on dense nonfiction—philosophy, law, medicine, and historical studies of the Edo period.
Store owners express ambivalence about the boom. While daily sales have multiplied fivefold in some cases, they recognize that removing these volumes from circulation permanently depletes their stock and could drive up prices for remaining copies. The practice mirrors similar operations already reported in the U.S. and Europe, raising fears that Japan's cultural archive is being quietly stripped for training data.
This practice could permanently remove significant portions of Japan's literary and historical record from public access, affecting scholars, students, and casual readers who rely on used bookstores for rare or out-of-print works. While AI developers may gain richer training data, the long-term societal cost could be a hollowed-out cultural archive. Future researchers may find primary sources only in fragmented digital scans, potentially altering how Japanese history and thought are studied.