What are we talking to - how should it talk

I noticed I couldn’t make Chatgpt stop apologizing - it is distracting and I notice this question which to me remains vague - what are we talking to and how should it generate intermediary statements like the general “As an AI model I cannot have an opinion” or “I apologize for apologizing.” It becomes generic quickly and breaks down the quality of the conversation. Another one is the constantly recurring structure of lists of suggestions or advice which keeps on happening even when not asking for it. What would be a good way to approach this issue?

One is that it is not a human being and empathy and support are good values yet they create a weird illusion talking to a human