Rare Books: Amazon shreds them to train its AI

The answer is simple, if a little brutal: to better teach artificial intelligence. AI models have already "read" almost everything available on the internet. To make a qualitative leap, they need unique, less common data that can't be found with a simple online search. This is where rare books come in.

These ancient, often unique volumes contain language, expressions, and knowledge that AI hasn't yet processed. Imagine texts in obsolete languages, specific dialects, or simply writing styles we no longer use. They offer a wealth of information that mainstream data cannot provide. Amazon, specifically, began destroying rare texts to train its LLMs, according to a TechCrunch report from August 17, 2026.

Don't imagine a library burning. The process is more scientific and, if anything, even colder. First, the books are digitized using high-resolution scanners. Every page, every illustration, every detail is captured and transformed into data. Once the information is "sucked out," the book, now useless for its purpose, is simply disposed of.

Sounds like a dystopian movie plot, doesn't it? The company born selling books now shreds them. This method ensures AI has access to a highly diverse text corpus, but at the cost of losing a physical piece of history. Isn't it a bit ironic, from a company founded by Jeff Bezos as an online bookstore, to see this approach?