I doubt that. Why would you train Mythos on its own code if you don't want it to be able to reproduce it? It's not going to add much to the overall corpus.