An article on roboethics is at the Times. In an ideal world, government authorities would recognize the friendliness problem in advance and pro-actively create Friendly AI with competent researchers, though I would currently estimate the chances of that happening in the next 20 years as less than 10%.