I think the only real use for AI today is with human oversight. Like, have it wrote your essay/code/business plan, but review it for actual use in the world. It can do something 75% accurate and add the knowledge of a competent human and you have 1 person doing the job of 4 people. It realistically 3 people since review takes time. This doesn't address the ethics of doing this, but is the only realistic way of using it in Mar 2024.
I absolutely know we'll deserve it, and we always have. That's one thing I'm sure Abrahamic religions get right, the other being that we can only progress as a species through helping and forgiving eachother(although the Romans also had the concept of forgiveness of debts as essential figured out).
I figure trying to be an exception to the "deserves it" rule is pointless, as since we are irrational and arbitrary beings, AI will undoubtably come up with countless criteria(both rational and less so) by which we deserve death that we can't fathom, let alone mitigate ourselves. Our best shot may be to teach AI the utility of forgiveness, along with whatever value categories we can tack on to the concept.