UK Booksellers Alerted as AI Firms Mass-Buy Rare Books for Scanning and Destruction
Bastillepost · 1 SOURCESabout 2 hours ago2 MIN
Summary
British second-hand booksellers are raising alarm over a mysterious surge in bulk purchases of rare and out-of-print books by American buyers. Evidence suggests these volumes are being shipped to US warehouses, stripped of their spines, digitally scanned, and then destroyed to serve as training data for artificial intelligence systems. Court documents and industry reports indicate Anthropic, the company behind chatbot Claude, has been at the forefront of this initiative through a program dubbed Project Panama. The revelation has ignited fierce debate over copyright, cultural heritage, and the ethics of AI development.
Key Points
- Lucy Hingston, manager of Compass Books in Devon, noticed unusual bulk orders covering disparate topics including Peruvian geology, South African history, railways, and Hong Kong jewelry
- Anthropic's Project Panama reportedly spent millions acquiring hundreds of millions of physical books for scanning and disposal
- The company previously trained models using over 7 million pirated ebooks, with 5 million sourced from a single piracy website, settling related litigation for $1.5 billion in August
- UK Publishers Association CEO Dan Conway criticized tech firms for unauthorized use of copyrighted material, saying it harms authors and creative industries
- Finn Windle, a staff member at a York bookshop, stated they would not have sold certain out-of-print titles had they known the intended use
Why It Matters
The systematic acquisition and destruction of physical books for AI training represents a significant threat to irreplaceable cultural records, particularly out-of-print works that exist in few copies worldwide. With online content already extensively utilized for AI training, this development signals that tech companies may increasingly turn to physical archives, raising urgent questions about copyright enforcement and the preservation of human knowledge in digital form.
The systematic acquisition and destruction of physical books for AI training represents a significant threat to irreplaceable cultural records, particularly out-of-print works that exist in few copies worldwide. With online content already extensively utilized for AI training, this development signals that tech companies may increasingly turn to physical archives, raising urgent questions about copyright enforcement and the preservation of human knowledge in digital form.