Booksellers suspect AI firms behind 'strange' bulk book orders
Secondhand bookshops across the UK and Ireland say mystery buyers are clearing their shelves with erratic, theme-free bulk orders — and the trade increasingly suspects the books are being scanned and pulped to feed AI training pipelines.
Booksellers in the UK and Ireland are reporting a wave of "strange" bulk orders from anonymous buyers, with many convinced the purchases are AI data acquisition. The Guardian spoke to shops fielding orders from the US, Canada, continental Europe and the UK — purchases that stand out for their randomness: one order mixed an Estonian translation of John le Carré's The Mission Song, a specific imprint of Anne Brontë's Agnes Grey, and the October 1983 issue of Warship magazine. Jim Shaughnessy of MW Books in Ireland said orders since May run from "books on agricultural implements in 18th century Africa through to biographies of 1950s racing car drivers." One anonymous UK bookseller has moved 6,000 books since January, with buyers paying "top whack" and no postcode trace beyond freight warehouses near Heathrow.
The suspected endgame: high-speed destructive scanning — cut the spine, feed pages through a machine, recycle the husk — to harvest the last unpolluted human text on Earth. 404 Media reported in July that the book database ISBNdb is actively pitching AI companies, promising "the world's best AI training data is sitting on a shelf" and coordinating purchases from 1,000 to a million books, all under strict NDAs. The pitch leans on a real technical fear: books published before 2022 predate chatbot-generated content, so they're structurally immune to the AI-slop contamination that drives model collapse. The practice already has legal cover — a federal judge ruled Anthropic's own book-scanning operation fair use, after the Washington Post revealed the company spent tens of millions slicing spines and scanning titles. Anthropic says none of its programs buy and destroy rare or antiquarian books.
The optics are the story now. Earlier this month Australian booksellers sounded the same alarm, and Irish shop Kennys has started refusing suspicious bulk orders outright. There's no confirmation of who the buyers are, which is exactly the problem: an industry that answers every data question with "we'll figure out the legalities later" is strip-mining the physical record of human knowledge to keep its models fed. Pulping a 1983 magazine is one thing; 18th-century agricultural manuals are another. The trade's patience is wearing thin, and one bad headline about a destroyed rare book could turn a quiet procurement pipeline into a full-blown copyright and cultural scandal.
What to watch: whether any lab steps forward to claim the buying spree, and whether ISBNdb's sourcing model survives the scrutiny.
If AI firms are quietly buying up and destroying old books for training data, where's the line between fair use and cultural vandalism? Tell us in the comments.
