The thing the headlines keep skipping over in pursuit of outrage-clicks is that this is how you scan books in bulk. The spine is removed and the pages are fed through a high-speed scanner in sequence.
And the books that these companies are scanning this way are not "rare" in the sense of Gutenberg bibles or first-editions of famous works. These are books that have been sitting untouched in libraries or warehouses for decades because nobody wants them and their alternative fate is simply to be pulped and recycled. AI companies aren't interested in quality, they're interested in quantity. It would indeed be nice if they'd release the results of doing these scans to the public, of course, but that's where copyright raises its head to ruin everything. They're just following the law, unfortunately.
I'm all for getting more books scanned and uploaded to Anna's Archive. But that's a separate issue.