What about Asimov's 3 laws of robotics?
The distinction is between:
Drone robots: Merely extensions of human control, which do what any other tool that extend human capabilities would do - including kill without remorse. Self-driving cars, aerial robots like Predator military aircraft, robotic solider machines with guns mounted on them or bombs built into them are all examples of drone robots that could or do easily kill.
"Thinking" robots - Which have programming which reproduces creative or deductive mental functions and have autonomous "intent". Building "intention" into a thinking machine is still something that has not been achieved except in very limited fashion. Programming that cannot modify itself is characteristic of a drone. A true thinking-machine would have programming that displays evidence of "intention" to how it processes data.
There are already computers which have the ability to process the 20 billion FLOPs or so that the human brain engages in, but human brains are associative neural networks (not linear circuits) in which there are an infinitely varied number of connections along that network that all feedback, interact with and modify the data-stream (so to speak). Computers are linear thinking machines because they use digital language which is simply a stream of numbers on an electrical circuit whereas human brains are chemical-electrical circuits with contributions by a whole host of neurochemicals and non-rational feedback mechanisms that "modify" the data stream.