AI Training Push Is Reportedly Destroying Rare Physical Books
A blog post from the operators of a shadow library argues that some AI companies acquiring large book collections for training data are using destructive scanning methods - slicing bindings apart to feed pages through automated scanners - and then discarding the physical originals afterward.
The post frames this as an urgent preservation issue, particularly for rare or out-of-print books where the scanned copy may become the only surviving version. It calls for coordinated efforts to digitize vulnerable texts more carefully, or at least in parallel with any destructive processes, before irreplaceable originals disappear.
The discussion on Hacker News centered on how AI training data pipelines treat physical media as disposable input rather than artifacts worth preserving, and what obligations (if any) buyers of bulk book lots have to libraries or archives.