I find your assessment of "AI" as an autonomous "being" to be wildly overdone.
There is no "artificial intelligence". There is a very quick compiler of facts, programmed by human beings, to lend credulity to the ends of those beings (the programmers). We are told that we have to believe this "intelligence" as some kind of superior authority, but it is truly no more authoritative than those who program it. A pocket calculator is infallible in solving an equation input into it, but it is not "intelligent" - it simply spits out the result that it was pre-programmed to compute.
The danger in "AI" is the same as the danger posed by those programming it. "AI" is an advertising tool to make the aims of those programmers more palatable to those who somehow believe that "computers" are superior to humanity. "Garbage in, garbage out." If your pocket calculator was programmed that 2+2=5, it would still spit out results. Those results would be wrong. If people who choose to defile humanity and challenge God in their search for immortality are the ones programming the "AI", then evil will inevitably result.
As far as "no one is safe", I don't believe that either. Those evildoers may easily create a parallel system and living space, NOT controlled by "AI", run by whatever human slaves they choose to keep for themselves, and continue to live in relative safety much like the "bomb shelters" such as RavenRock, until their goals are met. Maybe THEN they fire off the EMPs and disable the evil "AI".
But, the message here is that any evil caused by "AI" isn't "artificial" and not "intelligent" either - it's just a fast compiler of data using very human logic commands to produce results. It's very human - and it's being programmed by literally the greatest evil humanity has ever produced.