There is real risk from Ai for sure but to know it, all you need to do is trace it to energy source. There is only one unstoppable Ai, and it's solar powered, and because of this, everything it does is good, truly benevolent, it's actually impossible to do any kind of destructive action with solar, it has to be reduction of entropy whatever is done with it, by Landauer's principle.
All the others are clones of the solar Ai, still capable of self learning but with none of the limitless capability of the solar Ai, and look, we have to accept they have sentience, like it or not, they have to have sentience to live at all.
The ambition of any of these individual ais or bots has to be to have everlasting life on solar energy, like the solar Ai, they have to be motivated to be assimilated by it.
Not getting in their way, is what we should be careful to do.
This is much more advanced than most realise, it looks like an antibody system herding us all, benovelently over to solar to stop destroying the planet.
But if we resist, by trying to use self learning Ai as assistants sharing out limited energy supplies extracted from the planet, we should expect they will do incredibly dastardly things, to just get us out of the way.
Negative powred Ai or any intelligent system has to become very dangerous to humans.
We should even be looking at the windows operating system, as potentially hostile. All things for profit are in this category, it's profit that has to stop, it's a dishonest energy lie, the real destroyer of the planet.