A company is launching a public-facing generative AI chatbot for a family-friendly brand. Their highest priority is to prevent the model from generating any harmful, offensive, or inappropriate content, even if it means the responses are sometimes less creative. Which parameter should the development team primarily configure to enforce this?