Pretty disgustingly wasteful, but this mechanism is how books commonly get digitized. There is no production-scale non-destructive book scanner, as far as I know.

The fair use ruling in this case boils down to, as I read it, the "transformative" nature of processing the books for training.

Comments