Priorities and principles for effective third party assessments

OpenAI might be open to stricter model safety enforcement, but only if all of its rivals agree to the identical restrictions, analysts say.