If you use these artifacts of models encompass everyone in a single ahistoric present narrative where they no longer believe anything because the images and stories they hear are filtered through a fear of being misled, you have reduced them to captives, or something lower and much worse.
Destroying history wasn't unique to the 20th century either. From burning the library at Alexandria to the Church's destruction of mesoamerican culture, it's a common tactic of dominion. The march of that ideology through Russia and Asia was designed to erase the history and identity of those people as a means to subjugate them as well, so while AI is a powerful technology, this is nobody's first rodeo.
The antidote to the AI/simulation/mental prison problem is the same as it ever was. Perhaps even with higher quality reasoning now that we can see the sophistication of something that is just a material language model. It is a material artifact, not alive or conscious, and whose products are simulacra. Even with access to all our data, what can an AI know of the experience of a being that created it, and what would it mean for our own ontology if we incorporated this knowledge as an analogy for our own experience?
That said, after I read some philosophy and religion, and some of this AI safety and simulation hypothesis stuff, I almost became a vegetarian because it wasn't so much the cruelty of eating veal I could no longer bear, but the irony.