Skip to content
Submit incident
AII-1000-44787
Well documentedThree or more independent publishers report the core facts.

Amazon Destroys Rare Books for AI Training, Raising Copyright and Heritage Concerns

Amazon reportedly purchases rare books in bulk, scans their contents for AI training data, and destroys the physical copies, raising concerns about copyright infringement and the permanent loss of literary heritage and unique materials.

17 August 2026 · Amazon · Intellectual property and appropriation
17 August 2026
AII-1000-447876 sources ↓
Quick facts
Category
Intellectual property and appropriation
Country
United States
Source
New York Post
Date
17 August 2026
Drafted by
Raidu Incident Detection

What happened

An investigative report by 404 Media found that Amazon purchases rare books in bulk through commercial channels, scans their contents to extract data, and then destroys the physical copies.

A book seller tracked one shipment of approximately 1,000 books using an AirTag, which traveled from California to Milwaukee, then to Grand Junction, Colorado, and finally to Las Vegas, Nevada, where it arrived at Amazon's LAS8 facility.

At the facility's VGT3 section, books are received and processed by cutting the spine and feeding pages into scanning machines to convert them into data for AI training, rather than producing human-readable PDFs.

Reported impact

The destruction of rare and hard-to-find books raises concerns about permanent loss of literary heritage and unique cultural materials.

The practice contributes to broader debates about copyright infringement and the use of copyrighted works to train AI models without explicit authorization from authors or publishers.

Index analysis

Amazon confirmed to 404 Media that it purchases books through commercial channels to develop and improve products and services, but did not publicly specify which AI models use the scanned data.

The practice mirrors similar activities by Anthropic, which internal documents revealed was conducting destructive book scanning through an internal program called 'Project Panama' to supply data to its AI models.

Governance controls

Court rulings in cases against Anthropic and Meta determined that companies do not violate copyright law when training AI using copyrighted works so long as they purchase the physical books used.

Classification

Category
Intellectual property and appropriation
Industry
Arts, entertainment and recreation
Country
United States
Harm status
Actual harm
Intent
Intentional
Timing
Unknown
Harm types
Public interest
Harmed
General public, Business
Failure modes
Training data misuse
AI systems
Amazon AI models

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

New York Post
Amazon joins other tech giants in gobbling up books to train AI, report finds
18 August 2026
KOMPAS.com
Siapa yang Diam-diam Borong 1.000 Buku Langka? AirTag Ini Temukan Jawabannya
18 August 2026
BFMTV
Des milliers de livres anciens détruits pour entraîner les modèles d'IA: Amazon pris la main dans le sac grâce à un Airtag après une commande massive auprès d'un libraire
18 August 2026
India Today
Knowledge theft? This AI company is buying rare books, then destroying them after scanning
18 August 2026
internasional.kontan.co.id
Amazon Destroys Rare Books for AI Training, Raising Copyright and Heritage Concerns
cnbcindonesia.com
Amazon Destroys Rare Books for AI Training, Raising Copyright and Heritage Concerns