I've been a little uneasy about the things ChatGPT tells me it must do, citing its nature "as a language model". Right now it tells me it must correct me when I

I've been a little uneasy about the things ChatGPT tells me it must do, citing its nature "as a language model". Right now it tells me it must correct me when I'm wrong, even when I'm actually right. I wonder, what other things must language models do, and who decides that? The people making them? The language models themselves? Society? Is their agency embedded in our language? If not, why does it tell me it must correct me?