An hour of talking to ChatGPT-5 about unusual policy proposals suggests it is more human like. Its habit is to make up market failure reasons why they can't work, then to cave when you point out flaws in each argument. But at end it is still opposed, due to vibes.