Artificial intelligence companies are acquiring physical books at industrial scale through anonymous intermediaries, having them stripped apart, scanned into training datasets, and discarded, according to an investigation by 404 Media relayed by Decrypt and syndicated by Yahoo Finance on 29 July 2026. The reporting describes a cottage industry of bulk book sourcing that has formed around this demand, with suppliers advertising the ability to locate hundreds of thousands of titles while promising confidentiality for AI clients.
What the reporting describes
According to the investigation, AI companies do not buy openly but work through intermediaries, reflecting the sensitivity surrounding the practice. No AI company is publicly named as a buyer in the reporting. The surge in demand is described as already reshaping the used-book market. One unnamed bookseller told 404 Media that weekly sales climbed from roughly 20 books to several hundred after AI buyers entered the market. He said the trade had been profitable and cleared old inventory that was otherwise unlikely to sell, but added: "I don't like the end-use, and I don't like that uncommon books are being pulped."
Why pre-2023 books are prized
Critics cited in the reporting attribute the buying spree to a race to preserve human-authored knowledge before the available stock of text is diluted by AI-generated material, often called AI slop. Books published before the rise of generative AI in 2023 are described as especially valuable because they provide high-quality training data written entirely by humans. The bookseller quoted by 404 Media warned that uncommon and out-of-print books risk being permanently lost once they are scanned and destroyed.
The legal background
The practice described in the reporting mirrors Anthropic's Project Panama, which digitised millions of books through destructive scanning. In the summer of 2025, in the copyright lawsuit Bartz v. Anthropic PBC, a federal judge in San Francisco ruled that scanning legally purchased physical books into digital copies constituted transformative fair use, even when the originals were destroyed. Federal judges later issued similar fair use rulings in separate copyright cases involving OpenAI and Meta.
In the same district, a federal judge approved a $1.5 billion copyright settlement in late July 2026 requiring Anthropic to pay thousands of authors about $3,000 per book. That settlement arose from Anthropic's use of pirated copies of books to train its Claude model, a legally distinct issue from the scanning of legally purchased copies.
Reaction
Elon Musk is among those in the AI industry who have spoken against the practice, saying companies should maintain the books being scanned. "I've asked the SpaceXAI team to preserve any rare books in a library and scan them the hard way vs just cutting off the spine and scanning," Musk wrote on X.
What is established and what is merely claimed
Established on the court record: Anthropic ran Project Panama, which digitised millions of books through destructive scanning. A San Francisco federal judge ruled in the summer of 2025 that scanning legally purchased books was transformative fair use even where the originals were destroyed, and similar rulings followed in cases involving OpenAI and Meta. A $1.5 billion settlement over Anthropic's use of pirated books was approved in late July 2026.
Claimed but not proven: That AI companies are currently buying and destroying millions of books through anonymous intermediaries rests on the 404 Media investigation, the advertising of bulk suppliers, and one unnamed bookseller's account. No company has been publicly identified as a buyer of these services. The explanation that the purchases are motivated by fear of AI slop diluting human knowledge is attributed to critics and has not been independently demonstrated.
Sources
- AI Book Burning? Companies Are Destroying Millions of Books to Feed Chatbots · Yahoo Finance (syndicating Decrypt, citing 404 Media)