Any models trained on the ill-gotten data should now be public domain.