It would be impossible to train an AI on content from the public domain only? Impossible is an absolute, not buying it.