The ones that are “pretty darn good” are the ones that use the same underlying AI/ML tech as the average LLM, and would be in violation of this newly-formalized guideline.