What's more, it contained info on the suggestions it had issued to human reviewers regarding how to manage controversial topics, which includes that the AI need to "present to describe some viewpoints of people and movements", and never supply an argument "from its voice" in