OpenAI Board Forms Safety and Security Committee

OpenAI has established a Safety and Security Committee to oversee critical safety and security decisions for its projects and operations. This move comes as the company begins training its next frontier model, which OpenAI anticipates will significantly advance capabilities on the path toward Artificial General Intelligence (AGI).

Committee Composition and Leadership

The Safety and Security Committee is led by OpenAI directors Bret Taylor (Chair), Adam D’Angelo, Nicole Seligman, and CEO Sam Altman. To ensure technical depth, the committee includes technical and policy experts including:

  • Aleksander Madry: Head of Preparedness
  • Lilian Weng: Head of Safety Systems
  • Head of Alignment Science: John Schulman
  • Matt Knight: Head of Security
  • Chief Scientist: Jakub Pachocki

Additionally, Retired U.S. Army General Paul M. Nakasone, following his appointment to the Board of Directors, has joined the committee. The committee will also consult with external experts, including former cybersecurity officials Rob Joyce and John Carlin.

Immediate Objectives and 90-Day Review

The Safety and Security Committee's primary responsibility is to make recommendations to the full Board of Directors regarding critical safety and security decisions.

Its first major task is to evaluate and further develop OpenAI's existing processes and safeguards over a 90-day period. At the end of this window, the committee will present its recommendations to the full Board. Following the Board's review, OpenAI has committed to publicly share an update on the recommendations that are adopted, provided the disclosure is consistent with safety and security requirements.

Context of Frontier Model Development

OpenAI states that the training of its next frontier model is currently underway. The company describes these resulting systems as being expected to bring the organization to the next level of capabilities on the path to AGI. While OpenAI maintains that its models are industry-leading in both capabilities and safety, the company explicitly welcomes a "robust debate" regarding safety and security at this stage of development.

Sources