A shared playbook for trustworthy third party evaluations
OpenAI releases guidance for third-party AI evaluations. This playbook details how to assess frontier AI systems for capabilities, safeguards, and validity, promoting trustworthy AI development and deployment.