Are we just going to ignore that books are destroyed for AI?

AI companies are buying physical books in vast quantities, scanning them to train large language models, and destroying the originals by cutting off their bindings. The practice came to light through court documents in a lawsuit against Anthropic, which described “Project Panama”: a plan to obtain millions of books, scan them rapidly and discard them. Subsequent reporting suggests the buyers are not limiting themselves to common titles. Rare, antique, out-of-print and specialized books - some with no known digital copies - are also being acquired, raising concern among authors, archivists, historians and booksellers that valuable cultural material is being lost.

An investigation tracked a shipment of nearly 1,000 books from a rare-books seller to an Amazon-owned warehouse in Milwaukee, where workers reportedly scan books at scale. Companies turn to print because online book archives have exposed them to copyright lawsuits, while older, human-written books provide coherent, edited training material that is less likely to contain AI-generated text. Under current legal interpretations, buyers may destroy books they own, and scanning lawfully purchased copies for internal training has been treated as fair use. Critics argue this lets companies create digital datasets without paying creators. The total number destroyed is unknown: Anthropic documents projected scanning 500,000 to two million books in six months, while sellers report bulk orders ranging from hundreds to thousands of titles.