AI Book Burning? Companies Are Destroying Millions of Books to Feed Chatbots
AI companies are buying physical books at industrial scale through intermediaries, then stripping, scanning, and discarding them to build training datasets. The practice is driven by the need for high-quality human-written material, especially books published before generative AI became widespread. It is reshaping the used-book market, boosting sales for some sellers while raising fears that rare and out-of-print books are being permanently lost. The method resembles Anthropic’s destructive “Project Panama.” Courts have partly validated scanning legally purchased books as fair use, but Anthropic also recently faced a $1.5 billion settlement over use of pirated books. Backlash has prompted some AI figures, including Elon Musk, to urge preserving books and scanning them without destroying originals.
