Amazon's VGT3 facility destructively scans large volumes of books, workers say
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Amazon is now confirmed to be destructively scanning thousands of physical books at scale for AI training. This means your LLM’s next training corpus could include copyrighted or rare material without provenance, exposing you to legal risk and forcing you to either vet datasets more aggressively or accept higher compliance costs. If you’re fine-tuning or deploying models, expect tighter scrutiny of data sources and potential supply-chain audits.
Amazon is purchasing physical books in bulk orders of around 1,000 volumes to destructively scan them at dedicated facilities, bypassing digital scraping limitations to feed proprietary training pipelines. This aggressive shift to physical-to-digital data acquisition highlights that the LLM data moat is moving offline, meaning production teams relying on standard web-scraped open models will face a widening capability gap compared to hyperscalers with the capital to digitize physical libraries.