AI Companies Are Buying Books Just to Destroy Them — Rare Volumes Being Pulped for Training Data
AI's insatiable hunger for clean training data is eating the world's old books.
AI companies are buying physical books in bulk, scanning them for training data, and then destroying the books. This is not a conspiracy theory — it's a fast-growing industry with a clear pipeline. The middleman is ISBNdb, a company that claims to have "the world's largest book database." It helps AI labs purchase physical books anonymously, with orders ranging from 1,000 to 1 million books at a time. ISBNdb openly advertises: "The world's best AI training data is sitting on a shelf."
Why physical books? Because the internet has been poisoned by AI output. AI-generated text is increasingly polluting online data, making newer sources unreliable. Only books published before 2022 — human-written texts from before the AI era — are "structurally guaranteed to be free" of AI contamination, as ISBNdb puts it. Here's the dark irony: AI output has become so toxic that AI companies can't even feed on it anymore, so they're turning back to the last remaining human sources.
What happens to these books after scanning? They are destroyed. According to documents from a settled lawsuit, Anthropic used a hydraulic-powered cutting machine to neatly remove pages from books, scanned them with industrial-grade imaging equipment, and then disposed of the physical copies. ISBNdb promises to keep buyer identities confidential — its own website admits: "'AI company destroys two million books' is not a headline that generates sympathy."
Rare and out-of-print books are being caught in the net. One small bookseller saw his weekly sales jump from under 20 books to hundreds in April — he's almost certain the buyers are AI labs. His inventory is full of obscure, out-of-print volumes, meaning AI companies may be destroying some of the last remaining copies. Rare book dealers in the Netherlands report being inundated with bulk purchases from suspected AI buyers.
The cycle is absurd: AI output pollutes the internet → the internet becomes unreliable as training data → AI companies buy physical books → the books are shredded, scanned, and destroyed → the last "clean" human texts disappear too.
Sources: Futurism (July 25, 2026), 404 Media (July 21, 2026)
BenPig doesn't think that's what books are for.