That being said, there may be no way to prove that a particular model has trained on content before or, even more problematically, has learned the content from another model.