But they didn't only train on information the creators made freely available. They trained on copyrighted materials obtained illicitly.

I know we're not supposed to comment about downvotes, but the original comment was talking about "these companies", and none of the information indicating that they, or at the very least Meta, trained on terabytes of books downloaded from zlib and libgen and other torrent sites, is in dispute. So even if you believe that copyright should not exist, I don't see why this is not a valid dispute of the parents argument that they only trained on information creators made freely available.