Rare books and authors' work are at risk after 404 Media published reporting tying Amazon facilities to the practice of buying, scanning and destroying printed books for AI training. The outlet quotes its reporting as saying Amazon “is buying massive quantities of books, scanning them for AI training data, and destroying them in the process.”
The revelation lands amid a broader pattern. Journalistic and legal scrutiny has been building for more than a year over what critics call "destructive scanning" by AI firms. The Washington Post reported this January that Anthropic pursued a secret effort, dubbed Project Panama, to “destructively scan all the books in the world,” information that surfaced in a copyright lawsuit brought by authors earlier this year. Similar lawsuits now name Meta, Google, Microsoft and OpenAI, suggesting this is not limited to a single company.
404 Media traced one rare book's journey around the country and concluded the title ended at an Amazon warehouse in Las Vegas, Nevada. Employees at that location told the outlet they receive large shipments of printed books, remove bindings to speed scanning, and destroy the physical copies during the process. The reporting also describes the offending team’s logo as a dinosaur holding a book.
The reason AI firms prize printed books is laid out in documents The Washington Post reported on. According to those materials, one Anthropic co-founder theorized that training models on books could teach them ‘how to write well’ instead of mimicking ‘low quality internet speak.’ Emanuel Maiberg, writing for 404 Media, echoed that logic, noting, “Printed books are valuable as training data because a lot of the text they contain is not readily available on the internet, which AI companies have already scraped.” The documents also emphasize that books printed before 2022 are less likely to contain AI-generated text, a concern tied to a recursive degradation sometimes called 'model collapse.'
Amazon has previously denied or declined to acknowledge engaging in destructive scanning, and 404 Media presents its findings as signs of possible participation. The story intensifies existing legal and ethical questions about consent, copyright and cultural stewardship when commercial AI training relies on large-scale acquisition and destruction of physical books.
Public reaction could matter. Reporting and lawsuits have already pushed scrutiny onto corporate practices, and some companies have changed course after public pressure. The immediate next steps are likely more courtroom fights and further reporting into where physical texts are going and which teams are using them to build AI models. If you buy books through links on this site, Lit Hub may earn a commission from Bookshop.org, whose fees support independent bookstores.
