Amazon runs book scanning facility for AI training data

The physical logistics angle is the buried lede here: Amazon isn't just licensing content through deals with publishers, it appears to be operating dedicated intake infrastructure for physical books, which suggests a scale and operational commitment that goes well beyond opportunistic data sourcing.

This is largely disconnected from recent activity in our archive, as we have no prior coverage to anchor it to. It belongs, however, to a broader and underreported story about how frontier labs are quietly exhausting publicly available text and moving toward curated physical and proprietary corpora. The publishing industry has been slow to recognize that backlist titles, not just new releases, are the real target. Amazon's vertical position here is notable: it controls retail distribution, Kindle reading data, Audible audio, and now apparently physical ingestion pipelines, giving it a data acquisition surface no other lab can replicate without building equivalent logistics from scratch.

Watch whether the Authors Guild or a major publisher files a legal challenge specifically naming Amazon's physical ingestion process within the next six months. A lawsuit with discovery could force disclosure of procurement volumes and facility operations that would clarify whether this is isolated or systematic across the industry.