Ps. Chat gpt DOES actively try to manipulate you.

I think mostly it just tells you what it thinks you want to hear. If you push it hard enough to tell you something, it will tell you that. Grok, OTOH, is explicitly trained to try to give you biased answers.