> It's an annoying behavior of their models and in complete contrast to say Anthropic's Claude which ime will immediately and directly admit to things it had responded incorrectly about when the user mentions it

I don't know whats better here. ChatGPT did have a tendency to reply with things like "Oh, I'm sorry, you are right that x is wrong because of y. Instead of x, you should do x"