Models are not paranoid or anxious, they do not think or have feelings. I know you're probably using those words as a metaphor but we need to be careful about anthropomorphizing LLMs.

They are trained on natural language. Not anthropomorphizing them is the worse end of the spectrum.

They didn't describe the model, they described (accurately) the behaviour. They are useful descriptors of behaviour.

As an accelerationist and transhumanist, no way! These models passed the Turing test years ago. When a thing is indistinguishable from human, it is human. Our brains are, after all, just a collection of learned memetic weights. Just ask the determinists.