You would have to believe that an AI model would be 100% correct in its decision to discern an enemy from a civilian. So an intelligent lunatic, or an uninformed lunatic politician