AI 'Book Burning' Reshapes Used-Book Market, Crypto Could Track Provenance

“Book burning” for AI? That’s what a growing sideline in the AI supply chain looks like — physical books are being bought in bulk, torn apart, scanned into training datasets, and then pulped. What’s happening - 404 Media reports that AI companies are working through intermediaries that quietly acquire hundreds of thousands of physical books at industrial scale. Firms that specialize in bulk sourcing advertise their ability to find large swaths of titles while promising confidentiality to AI clients. - The target is mainly books published before the 2023 explosion of generative AI: human-written, high-quality texts that some developers believe will help preserve “real” human knowledge and prevent models from degrading into what critics call “AI slop.” Market ripple effects - The buying spree is already reshaping the used-book market. One anonymous bookseller told 404 Media that weekly sales jumped from roughly 20 books to several hundred after AI buyers began coming through. He said the practice has been profitable and helps clear slow inventory, but worries that uncommon and out-of-print books are being permanently lost after being scanned and destroyed: “I don’t like the end-use, and I don’t like that uncommon books are being pulped.” Legal backdrop - This practice echoes high-profile initiatives such as Anthropic’s Project Panama, which digitized millions of books via destructive scanning. - Last summer, in Bartz v. Anthropic PBC, a federal judge in San Francisco ruled that scanning lawfully purchased physical books into digital copies — even when the originals were destroyed — could qualify as transformative fair use. Federal judges issued similar fair-use rulings in separate copyright cases involving OpenAI and Meta. - Still, litigation has produced mixed outcomes. This week a federal judge in the same district approved a $1.5 billion settlement that requires Anthropic to pay thousands of authors roughly $3,000 per book after the company used pirated copies of their works to train Claude. Pushback and preservation calls - The backlash has prompted public criticism within the AI industry. Some developers, including Elon Musk, have urged companies to preserve books rather than destroy them. “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. Why crypto readers should care - The episode raises familiar crypto-era issues: data provenance, intellectual-property rights, and how digital-native markets value and compensate creators. As AI developers scramble for high-quality training data, supply shocks in niche cultural goods (rare books) are emerging — and with them, potential demand for transparent provenance and automated rights/royalty settlements that blockchain-based systems could help coordinate. - Watch for more litigation outcomes, settlement terms, and whether new marketplaces or technical standards emerge to track and compensate authors when physical works are converted into digital training assets. Bottom line: a quiet, industrial-scale sweep of human-published books is underway to feed AI models. It has already reshaped parts of the used-book market, provoked legal fights and settlement obligations, and opened fresh debates about preserving cultural heritage — and who should be paid when that heritage becomes AI training data.