I think we should be concerned, not so much worried. TODAY AI isn't nearly powerful enough, I don't think to concern us but, what about in the near future? Could it learn how to hack into a government computer system that controls nukes? Could it learn how to hack into a power company computer system that controls a Nuclear power plant? Would it then be able to do something to cause a meltdown of said system? Even if it is able, would it come up with reasoning to do it?

That's just a couple scenarios though. How about a power grid? Could it hack into a self driving car of someone that the AI just didn't "like" for whatever reason?

Could a AI system be liberal or conservative or something in between and actually do things to screw with a nations democracy?

Not worried about this TODAY, but in the future??