I agree, it seems like there should be a traditional program on top that's filtering responses for known company secrets, conversations that go against published company guidelines, etc.