Safety and Security Measures Update

OpenAI is converting its Safety and Security Committee (SSC) into an independent board oversight committee to oversee key safety and security measures throughout model development and deployment. This decision reflects the results of a 90-day review conducted by the committee and the recommendations that followed.

The new committee will be chaired by Zico Kolter of Carnegie Mellon University, with Adam D’Angelo, Paul Nakasone, and Nicole Seligman participating. The committee will receive safety evaluation reports before major model launches and will have the authority to delay model releases until safety issues are resolved.

The key enhancements are as follows:

• Establishing independent governance: Setting up an independent oversight structure under the board
• Strengthening security measures: Expanding cybersecurity capabilities and considering the establishment of an Information Sharing and Analysis Center (ISAC) within the AI industry
• Improving transparency: Disclosing model capabilities and risks through System Card
• Collaboration and framework integration: Collaborating with external organizations and building an integrated safety framework for model development and monitoring