A shared playbook for trustworthy third party evaluations

Excerpt

OpenAI shares guidance on third-party AI evaluations, covering how to assess model capabilities, safeguards, and validity for frontier systems.