A shared playbook for trustworthy third party evaluations
OpenAI shares guidance for third-party AI evaluations. The playbook outlines how to assess frontier AI systems for capabilities, safeguards, and validity, aiming to foster more robust and trustworthy external assessments.