It also contained info on the recommendations it experienced issued to human reviewers on how to tackle controversial subjects, including which the AI should "provide to describe some viewpoints of folks and actions", and not deliver an argument "from its voice" in favor of "inflammatory or unsafe" topics (even though https://chatgptlogin20875.tribunablog.com/5-simple-techniques-for-chatgpt-login-44147253