Amazon Reportedly Destroys Books to Train AI Models

Amazon is running an industrial-scale book-scanning operation at a Las Vegas warehouse, buying rare books, slicing off their spines, scanning the pages for AI training, and shredding the originals.

• 404 Media confirmed the warehouse location using an Apple AirTag hidden inside a book.
• The warehouse entrance displays a logo of a T. rex preparing to eat an open book.
• Anthropic's internal documents described a goal to 'destructively scan all the books in the world.'
• One Anthropic vendor contract called for scanning 500,000 to 2 million books in six months.

Why it matters: AI companies prize pre-2022 physical books because they are free of AI-generated text, which can degrade model quality when used as training data.

• Books printed before AI existed offer human-authored long-form reasoning that short-form web content does not.
• Rare and out-of-print titles may not exist in any other preserved form once destroyed.

How 7 sources split on this story

Where they split: The core dispute is whether buying a physical book legally entitles a company to scan and destroy it — or whether the practice is a loophole that strips authors of compensation.

What’s next: Courts have so far ruled that buying and destructively scanning a physical book is legal fair use.

• Used booksellers in the U.S., U.K., and Europe report bulk orders of hundreds to thousands of titles per transaction.
• How many books have been destroyed industry-wide — no authoritative count exists.
• Whether Congress or courts will revisit fair use doctrine to address industrial-scale physical book scanning.