Amazon has been quietly buying rare books to feed its AI training models, and according to reports, the company destroys the original volumes once they've been digitized. The practice, which has drawn criticism from preservation advocates and authors, raises uncomfortable questions about how far tech companies will go to secure data.
The Reported Practice
Details are thin, but the outline is clear: Amazon acquires out-of-print and scarce books, scans them, and then discards the physical copies. Sources familiar with the process say the destruction is routine, though Amazon hasn't publicly acknowledged it. The company has not responded to requests for comment.
Why destroy the books? The reasoning, as described by people in the rare book trade, is to ensure the digitized version is the only one that exists — a way to maintain exclusivity over the content. That logic makes sense for AI training, where having unique data can give a model an edge. But it's a brutal one for anyone who cares about the survival of physical texts.
Concerns Over Cultural Preservation
Rare books aren't just containers for words. They're artifacts — with provenance, marginalia, binding, and history baked into the paper. Once they're gone, they're gone. Preservationists argue that even if a digital copy exists, it doesn't capture the full object, and it certainly doesn't allow future researchers to study the original.
There's also a legal and ethical gray area. Amazon isn't stealing the books; it's buying them from dealers or auctions. But turning a one-of-a-kind item into training data and then destroying it feels different from scanning a library book. It's a permanent loss, not a temporary loan.
The broader worry is about the future of human-authored content. If AI models are trained on the works of living writers and then generate text that mimics those styles, what happens to the incentive to write? Authors have already sued tech companies over unauthorized use of their books. This reported practice takes that concern a step further — it's not just using the words, but erasing the physical evidence.
Some in the publishing world see this as a land grab. By controlling the only copy of a text, a tech giant could effectively own the intellectual legacy embedded in it. That's a different kind of monopoly than just owning the copyright.
No Clear Answer Yet
Amazon hasn't said whether it plans to change course. The company's silence leaves a lot of open questions: How many books have been destroyed? What criteria determine which ones get the treatment? And is there any oversight?
For now, the rare book community is watching. Some dealers say they're rethinking who they sell to. Others are asking whether Amazon's practice is even legal, given that some books may be under copyright protection. That's a question likely to land in court before long.
