OpenAI Plans to Advance Third-Party Safety Assessments to Model Development Stage
OpenAI plans to allow external organizations to conduct technical safety reviews during model training, evaluation, and release, advancing the timing of third-party assessments.