Booksellers suspect AI companies are purchasing rare books to destructively scan them for training data, raising concerns about the permanent loss of irreplaceable physical copies. While non-destructive scanning technology like Google's 2009 patent exists, cost and speed pressures in the AI race may be driving firms toward the cheaper destruction method instead.
Background
AI companies have increasingly turned to book collections for training data, with Google Books partnering with libraries since 2004 to digitize millions of titles. The Internet Archive and other organizations have pioneered preservation-focused scanning methods that prioritize keeping physical copies intact.
- Source
- Ars Technica
- Published
- Aug 12, 2026 at 11:19 PM
- Score
- 6.0 / 10