It's not. If you actually look at any chain-of-thought stuff long enough, you'll see instances where what it delivers directly contradicts the "thoughts."

If your AI is *ist in effect but told not to be, it will just manifest as highlighting negative things more often for the people it has bad vibes for. Just like people will do.