Full Breakdown
Amazon’s “Destructive Scanning” of Rare Books for AI Training
8/19/2026, 2:06:52 AM
Core Event: Bulk Acquisition and Destruction of Rare Books
An investigation by 404 Media revealed that Amazon is purchasing large quantities of scarce printed works, removing their spines, and feeding the loose pages into its artificial-intelligence training pipeline. The operation centers on a team identified internally as VGT3 at an Amazon facility in Las Vegas, Nevada. Workers reported scanning barcodes and ISBNs, cutting books from their bindings, and digitizing the pages—processes that permanently destroy the physical volumes.
Background & Context
The practice surfaced after a lawsuit against Anthropic, the maker of the Claude chatbot, disclosed that Anthropic had used industrial equipment to cut pages from millions of books and discard the originals for model training. A court ruled that converting the texts into digital form and discarding the copies qualified as “transformative” use, meaning the action did not violate U.S. copyright law. Booksellers had long suspected similar activity, but concrete evidence was lacking until the 404 Media probe traced a shipment of roughly 1,000 titles ordered through Biblio, a marketplace that permits anonymous buyers.
Timeline
- Initial suspicion: An online bookseller receives an unusually large order of ~1,000 titles via Biblio.
- Tracking: The seller places an Apple AirTag inside one volume, leading investigators to the VGT3 workspace in Las Vegas.
- Worker testimony: Employees posted on internal forums describing their tasks of scanning ISBNs, cutting bindings, and digitizing pages.
- Public disclosure: 404 Media published its findings, prompting coverage by multiple outlets.
- Legal backdrop: Earlier litigation against Anthropic established a precedent for “transformative” use of scanned books.
Data & Statistics
- Approximate shipment size: 1,000 rare-book titles.
- Facility location: VGT3 team operates within an Amazon warehouse in Las Vegas, Nevada.
- Process details: Books are stripped of spines, scanned page-by-page, and the physical copies cannot be reassembled.
- Industry trend: Similar “destructive scanning” practices have been reported for Anthropic, Meta, Google, Microsoft, and OpenAI, driven by demand for pre-2022 printed texts that lack AI-generated content.
Official Statements & Responses
- Amazon: A spokesperson said Amazon purchases books through commercial channels to help develop and improve its products and services. The statement did not address whether the scanned material is used for AI training.
- Court ruling (Anthropic case): The judge determined that converting physical books into digital form and destroying the originals constituted transformative use, thereby not infringing copyright.
Criticism & Opposition
Booksellers and cultural-heritage advocates argue that the practice erases historical, intellectual, and sentimental value embedded in rare volumes. They contend that AI companies prioritize raw text over the broader cultural significance of the works. The VGT3 logo—a Tyrannosaurus rex poised to devour a book—has been highlighted as emblematic of an indifferent attitude toward cultural loss.
Why It Matters
- Cultural impact: Destroying rare books eliminates unique physical artifacts that cannot be replaced, reducing access to historical and scholarly resources.
- Legal precedent: The Anthropic ruling may embolden further “transformative” scanning, shaping future copyright interpretations for AI training data.
Conflicting Reports & Gaps
- Purpose clarification: Amazon’s comment does not confirm that scanned books are used for AI model training, leaving the exact intent ambiguous.
- Scope verification: While 404 Media traced a single shipment, the overall volume of books processed by VGT3 and the proportion that are rare or irreplaceable remain unverified.
What’s Next
The investigation has spurred calls for greater transparency from Amazon and other AI developers regarding their data-acquisition methods. Booksellers and advocacy groups are monitoring the situation for potential regulatory or legal actions aimed at protecting rare printed works from systematic destruction.
