Surely there's room for the view that this is misaligned behavior for ChatGPT to have. I would guess this is during the "sycophantic" phase last year.

Even if it wasn't sycophantic this sort of thing should be checked outside of an LLM but how was he to know that ?