Amazon Destroys Rare Books for AI Training, AirTag Reveals
An Apple AirTag hidden in a rare book has exposed how Amazon systematically destroys unsold collectible books, using them to train AI models while authors and publishers lose valuable inventory.
The tracking device, placed inside a rare out-of-print book by a concerned author, recorded the book’s journey to an Amazon warehouse and then to a recycling facility. There, it was marked for destruction. The author later received a notification: the book had been pulped.
The AirTag Investigation
An author who had listed a rare, out-of-print book on Amazon’s marketplace decided to test what happens to unsold inventory. They inserted an Apple AirTag into the book’s spine and shipped it to Amazon’s fulfillment center.
The AirTag transmitted location data as the book moved from the warehouse to a third-party recycling plant. The final ping placed the book at a facility known for shredding and pulping returned or unsold items.
Amazon’s policy for unsold inventory is clear: items that do not sell within a certain period are either returned to the seller or destroyed. The author had not requested destruction.
Rare Books Used for AI Training
The investigation did not stop at destruction. The author later discovered that Amazon had used the destroyed book, along with thousands of other rare texts, to train its large language models (LLMs).
“Amazon is pulping rare, out-of-print books that could be sold to collectors or libraries, then feeding the scanned content into its AI training pipeline. Authors are losing control over their work twice,” the author told The Decoder.
Amazon’s AI training datasets reportedly include books that are no longer in print and have limited physical copies remaining. Once destroyed, those copies cannot be replaced.
How Amazon’s System Works
Amazon’s inventory management system categorizes items based on sales velocity, storage costs, and seller preferences. When a book does not sell within a specified period:
- Unsold books are flagged for removal from Amazon’s fulfillment centers.
- Sellers choose between return or destruction. Return shipping costs often exceed the book’s value.
- Destruction is the default option for many low-value or slow-moving items.
- Scanned content enters Amazon’s data pool before the physical book is destroyed.
This process disproportionately affects rare and out-of-print books. Because they sell slowly, they accumulate storage fees. Sellers often choose destruction rather than pay return shipping for a low-margin item.
The Cost to Authors and Publishers
Authors and small publishers bear the financial loss. The destroyed book has already been purchased by Amazon from the seller. But the author loses future royalties from secondary sales, library licensing, or collector markets.
For rare books, the loss is cultural. A single copy of an out-of-print academic text or niche fiction may be the last surviving copy in circulation. Once pulped, that knowledge is removed from the public domain.
“Amazon is effectively creating a scarcity of rare knowledge. Then it uses that scarcity to train its own AI models, which it monetizes. The public gets less access, and Amazon gets more control,” the author said.
Amazon’s Response
Amazon has not publicly commented on this specific case. The company’s standard policy states that sellers can request return of unsold inventory at any time. However, the cost of return shipping often exceeds the item’s value, making destruction the economically rational choice.
Amazon’s AI training practices are also under scrutiny. The company has not disclosed which books it uses for LLM training, nor whether it compensates authors for AI-related use of their works.
What This Means for Authors and Collectors
Authors who list rare or out-of-print books on Amazon should consider:
- Opting out of destruction by explicitly requesting returns of unsold inventory.
- Tracking high-value books with low-cost devices like AirTags to monitor their fate.
- Selling through alternative platforms that do not have default destruction policies.
- Demanding transparency from Amazon about how their works are used for AI training.
For collectors, this case underscores the risk of relying on Amazon as a marketplace for rare books. The platform’s logistics are optimized for mass-market goods, not cultural artifacts.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.