In Star Trek The Next Generation and Knight Rider are embodied examples of your problem. In each case there are two successful self thinking completely antonymous machines. One respects law and order and the other does not even think there is law and order. Like you I believe in half of one and half the other. You can't protect real living things from artificial life without limiting the artificial potential of these machines. You can however program a non erasable conscience that follows law and order.
RE: Should We Be Afraid of Artificial Intelligence?