Given the disasters that regularly recur in IT by government, I firmly believe they should be kept out of it. Despite all the hoopla, machine learning isn't AI and, frankly, isn't ready for prime time, especially in anything resembling safety-critical or other life threatening issues. Even financial issues, perhaps especially financial issues can be life threatening when machine learning is applied without absolute assurance that the models and training results in correct behavior. We need less "Black Swans" not more.