SkyNet was to be fiction, but some see it as plausible. I personally think AI will “Blue Screen” before it gets anywhere, but the idea is intriguing. Can a “bug free” AI emerge with a “personality” with designs on society? The remote possibility is if you combine two concepts in a militarized computer program: siege and perimeter control. You could go further but such possibilities first require pursuit by people. Essentially, you have to limit regulatory software and keep back temptation to fully automate all decision-making.