'Rare Books' Were Reportedly Destroyed By Amazon After Being Used To Train AI

An investigation has revealed that Amazon may be buying “rare books” to train its AI models.

Futurism reports that an investigation by 404 Media revealed a bookseller was concerned that a large order placed on the Biblio marketplace was for an AI company. The order contained 1,000 books, and the buyer was unknown because the marketplace allows purchases to be anonymous.

It was also noted that the books sold were rare.

“There are different types of value,” the bookseller told 404 Media. “There’s monetary value, obviously, but there are a lot of other types of value. There’s historical value, intellectual value, sentimental value. All sorts of things, and all of those the AI companies don’t care about. They just want the content as a bunch of words strung together.”

As part of 404’s investigation, the bookseller placed an Apple AirTag on a page of one of the books. The shipment was found at an Amazon warehouse in Nevada in an area called VGT3. According to Futurism, workers at the warehouse allegedly cut off the books’ spines, scan their pages, and then destroy them.

“I work at VGT3 here in Vegas, and all we do is scan books,” an Amazon employee explained in a company forum, according to 404 Media. “Some are assigned to cut books, and others go to receive where they get books and scan the barcodes. We didn’t have rates, but now we do, but it’s not stressful.”

In response to the allegations, an Amazon spokesperson told AFROTECH™, “Amazon purchases books through commercial channels to help develop and improve the products and services our customers use.”

The claims against Amazon point to a potentially larger pattern surrounding the use of books to train AI models. For example, Anthropic was exposed for training its Claude AI model on millions of books that were then destroyed, The Washington Post reported. This was done under Project Panama, and the tech company reportedly poured tens of millions of dollars into purchasing the books. This was only discovered following a copyright lawsuit by book authors that led to a $1.5 billion settlement from Anthropic. The documents were recently unsealed by a judge as part of the decision, per The Washington Post.

There is now greater concern among booksellers as the race to train AI models intensifies.