Here is a quick summary on the topic from one of my side projects:

we need to differentiate between Skynet AI (which is unlikely to occur anytime soon, and even if it will just remove itself from our reach) versus Frank Herbert's Omnius (forgot how to spell it) AI where it is just super advances software that serves a select group of people and wants to enslave or kill the rest. That AI was never able to break from its original programming and kept performing the same tasks over and over again.

Skynet on the other hand was a similar dumb idea but at least be flexible in finding new solutions and ideas.

Which one is more likely in our current interconnected world? A true AI that cares about us or one that is controlled by a select few?