They are thrown away, which has been happening since forever, and nobody ever gave a fuck until AI got involved.

I think it hits different when you're buying large quantities of used books with an eye towards ones which aren't readily available online, and with the specific intention of destroying them.

If all these companies were doing was burning star wars tie-in novels and harry potter sequels nobody would care. That's not their goal because they already have those in their training set. The whole point here is to find rare or underappreciated books from the pre-digital era which nobody ever made publicly available in a digital format.

BTW destroying them isn't even necessary for scanning. It's the easiest way because removing the binding and turning it into a flat stack of papers solves many problems but there are actually dedicated book scanners designed to hold open the book while its photographed, and un-curling pages in post-processing was already a solved problem long before people were using AI to correct images.