AI Companies Bulk-Buying and Destroying Rare Books to Train Language Models

Executive Briefing

• Reveals AI firms purchasing physical books in bulk, scanning contents, then destroying originals to bypass copyright restrictions
• Exploits first-sale doctrine, with courts ruling physical-to-digital conversion 'transformative' and protected under fair use
• Notes one bookseller's weekly sales surged from 20 to hundreds, with rare and out-of-print titles at risk of permanent loss
• Highlights broker ISBNdb facilitating million-book orders while keeping AI company buyers anonymous to avoid public backlash
• Warns pre-2022 print books are prized for being free of AI-generated text contamination, accelerating destruction of older titles