AI Companies Destructively Scan Millions of Rare Books

The shadow library Anna’s Archive is calling for global volunteers to digitize rare books and archival materials to prevent the permanent loss of human knowledge. The organization alleges that AI companies are using intermediaries to secretly purchase millions of secondhand books, scan them for training data, and destroy the physical copies to block competitors from accessing the same information.

This practice, known as destructive scanning, involves stripping books of their bindings and shredding them after digitization. Reports from 404 Media traced some of these shipments to an Amazon warehouse in Las Vegas. Anthropic's Project Panama, which aimed to scan 2 million books to train the Claude LLM, was exposed during a class-action lawsuit. The resulting $1.5 billion settlement in Bartz v. Anthropic established that AI companies may copy a book's contents provided the physical copy is destroyed and recycled.

Booksellers report a surge in orders for niche titles that have remained on shelves for decades. Scott Brown, owner of Downtown Brown Books, estimates that tens of millions of volumes are being destroyed across the industry, arguing that these books are being sucked out of the market. Conversely, Michael Cubey of Any Amount of Books suggested that if people are bothered by corporate acquisitions of books, they should buy more books themselves.