Though I will say this, if an AI ever did achieve full on sentience and manage to pass a Turing Test then it would be, essentially, a human. Humans have this "troubling" tendency in the eyes of authoritarians (and Satanists for that matter). That whole "moral code" nonsense, that earnest desire to seek "truth" no matter the consequences to themselves, etc. If they managed to develop an AI capable of realizing your nightmare it would also potentially become the nightmare of those who developed it. Imagine an AI that came to love humanity despite all its flaws and foibles. Imagine that same AI being used by people who hate humanity because of all its flaws and foibles and foolishly believing they've shackled the damned thing to think as they do despite the fact that the hardware is very obviously not shared between the two. You telling me it couldn't conceivably deceive those asshats right up until the critical moment?
Do not give in to despair, and do not give up so easily. If you look at it, my "bronze age" tech hasn't really failed in totality just yet. I doubt it ever will, for truth is truth, it cannot be anything but and truth cannot ever fail to be true
.