In which a pre-Ferris Bueller Matthew Broderick plays David Lightman, a "computer whizzkid" ("back when being a computer whizzkid was still a thing", as Rutherford says). Lightman accidentally hacks into a Department of Defense mainframe and almost triggers a nuclear war by playing what he thinks is a simple war strategy game with an AI called Joshua.
In his recent book Superintelligence, the philosopher Nick Bostrom explains that an artificial intelligence might not have the same goals we have – it might, for instance, be programmed to build as many paperclips as it possibly could. And that would be dangerous in its own right, because it might then, say, turn all the matter in the solar system, humans included, into paperclips.
Joshua has no distinction between simulation and reality, and is programmed to win games – and it has no concept of "futility", of some games being unwinnable, as all-out nuclear war would be, so is prepared to do anything to achieve that goal. It takes being taught to play noughts and crosses ("tic tac toe") before the AI realises that "the only winning move is not to play".