OpenAI Establishes Independent Committee: “Safety Board”

Martin Richter September 21, 2024 mysecurityevent Blog
OpenAI Establishes Independent Committee: “Safety Board”

OpenAI has established a new independent body, the “Board Oversight Committee”, to oversee the company’s safety practices. This committee has the authority to halt the release of AI models for up to 90 days if safety concerns arise. The establishment comes against the backdrop of growing concerns regarding the safety of Artificial Intelligence and internal conflicts within the company.

Key Insights

  • OpenAI is transforming its existing safety committee into an independent body.
  • The “Board Oversight Committee” has the authority to stop releases of AI models.
  • Members include well-known AI experts such as Adam D’Angelo and Nicole Seligman.
  • The independence of the committee is questioned, as the members also sit on the board of OpenAI.
  • The establishment comes against the backdrop of internal conflicts and growing safety concerns.

Background of the Establishment

The establishment of the “Board Oversight Committee” is a response to the increasing concerns regarding the safety of AI technologies. OpenAI has previously struggled with internal conflicts, which even led to the temporary dismissal of CEO Sam Altman. Several senior scientists left the company, highlighting the need for an independent body to oversee safety practices.

Composition of the Committee

The new committee consists of renowned experts in the field of Artificial Intelligence:

  1. Adam D’Angelo – Former CTO of Facebook and a member of the board of OpenAI since 2018.
  2. Nicole Seligman – Lawyer with experience at Sony.
  3. Paul Nakasone – Former US Army General and cybersecurity expert.
  4. Zico Kolter – Professor at Carnegie Mellon University and head of the committee.

Independence and Challenges

Although the committee is designed to be independent, the question remains how independently it can actually operate, as the members also sit on the board of OpenAI. The committee is regularly informed about developments at OpenAI and has a 90-day deadline to intervene in case of safety concerns. Additionally, new models must be submitted to US authorities for review before they are released.

Comparison with Other Companies

Similar steps have also been taken by other tech giants. For example, Meta has established an Oversight Board that makes independent decisions on controversial content on Facebook and Instagram. Unlike OpenAI, the members of this board are not involved in other roles at Meta, which strengthens its independence.

Conclusion

The establishment of the “Board Oversight Committee” by OpenAI is an important step towards greater transparency and accountability in the handling of Artificial Intelligence. It remains to be seen how effectively the committee will operate in practice and whether it will indeed be able to ensure the safety of the developed AI models.

Sources