AI companies are bulk buying and destroying rare books to train their models, alarming booksellers and book lovers. The cheapest scanning method involves cutting spines and discarding pages, risking permanent loss of irreplaceable texts. Google patented a non-destructive method in 2009, but AI firms prioritize speed and cost. The Internet Archive, led by scanner Eliza Zhang, has scanned over 3 million pages carefully by hand, ensuring zero errors and preserving fragile volumes. Zhang uses a foot pedal, adjusts cameras, and re-scans any blurry pages—a process that contrasts with AI firms’ rush.

Rare booksellers now flag suspicious orders, like Kennys Bookshop in Ireland receiving a “bananas” order for 5,000 obscure titles with no price haggling—a red flag for AI involvement. Tomás Kenny of Kennys said AI is “terrifying our industry” and plans to refuse such orders. Meanwhile, Anthropic was outed for destroying millions of print books in 2025 (though they deny harming rare books). Some AI firms, like OpenAI and Microsoft, work with Harvard to preserve public-domain texts. Yet critics doubt promises, as Musk’s xAI only pledged to avoid destroying “rare” books—a vague term. Booksellers fear that selling to destructive AI firms could harm both authors’ rights and reader access to hard-to-find titles.