"I'm sorry Dave, I'm afraid I can't do that."
A terrifying response if you need Hal to open the pod bay doors, but one that researchers at Tuft University, Massachusetts, believe will make robots safer to work with.
The research team isn't teaching their robots to behave with the emotional depth of a hormonal teenager - their response of "No" to certain requests stems from the robot's ability to assess an instruction and decided if it should carry it out.
For example, the instruction of "Go over there" might involve a robot putting itself at risk (if "there" is off a cliff). As such, the robot is able to reply in the negative before explaining why it has made that decision "There's no support ahead".
It could prove to be a very helpful mechanism that actually keeps humans safe: say you're on a demolition site, tearing down a building with a robot. The request "Knock down that wall" might get the response "No. That's a supporting wall structure. It would collapse and cause us serious injury" - which is a great deal more helpful than a robot blindly going about your bidding.
Not that we'll be sleeping any easier at night knowing that robots are now able to make their own decisions.
Terminator is starting to look less like fiction, and more like a documentary.