I've used AIs before, they might pose a threat in a distant future but it's having a tough time imitating us and it's just imitating, it can't understand anything, it just uses factors to determine probabilities based on past experiences and behaviors.
I must admit that reading Nick Bostrom's "Superintelligence" raised some concerns.
But on the other hand, I don't really believe we will develop sentient AI in the near future.
I think more likely humans will morph with AI long before we develop independent sentient AI.
It already is - AI driven suggestion features alone, and the ability for foreign actors to easily game them to push propaganda, have driven an insane amount of political instability. Who knows what the future can bring.