All LLMs are like this, just some are to a different extent. It's because of the fine-tuning step. Personally I want to fine tune a model to be a real shithead every once in a while. Would be refreshing.