For AI to attempt to destroy humanity using a nuclear war or other destructive method like that it'd need to be fully established with it's data centres (where it lives) up in space, away from all the destruction. Not only that, it'd need a self-repairing infrastructure up there too, with autonomous, self-creating/repairing robots to keep everything running smoothly. So, don't panic humans...your advance next word generator isn't going to kill you...just yet!! But once Elon has built his data centres up in space and Tesla has built the robots, maybe then our news should be discussing how dangerous AI 'could' be.