Unintended consequences certainly gets more and more dangerous the more advanced autonomy these weapons gets. In keeping with your mention of sci-fi, there's a Star Trek Voyager episode where two sides of killer robots continue to wage war against each other after having killed their makers because their makers had agreed on a truce, and their instructions had been to eliminate anyone that prevented them from beating their enemy and hence by agreeing a truce their makers had become an obstacle to achieving their programmed goal.

Having done software development my entire life, I'd never trust anyone to manage to avoid every pitfall in guaranteeing you have a foolproof way of making them stop...