OpenAI plainly admitted that it is impossible not to do so in a House of Lords inquiry. So, presumably there is no way around it to train models. There is just not enough non-copyrighted data out there.

I thought the outcome of that was basically it's legal to train on books, but they acquired the books in the wrong way. If they went out and bought copies of them and trained it would have been fine

Of course they are. They have just put on their Swiss Banker suit now and have all sorts of deflection techniques in place such that, of course, "the money has the stamps that says its clean" (when it it really blood money hidden behind a pretty wall).