It will take 90 days for OpenAI’s new Security and Safety Committee to assess and create security measures.
OpenAI announced the appointment of Sam Altman as Chair of the Safety and Security Committee, along with Bret Taylor, Adam D’Angelo, and Nicole Seligman as OpenAI Board Directors. The committee will be responsible for recommending safety and security choices for OpenAI initiatives to the entire Board.
Aleksander Madry, the Head of Preparedness at OpenAI, Lilian Weng, the Head of Safety Systems at OpenAI, John Schulman, the Head of Alignment Science at OpenAI, Matt Knight, the Head of Security, and Chief Scientist Jakub Pachocki, who recently succeeded Ilya Sutskever, will also be on the committee as technical and policy experts. Jon Carlin and Rob Joyce, two former cybersecurity officials, will also be consulted.
It’s interesting to note that the announcement coincides with the confirmation that work on the next better model is ongoing. During the most recent Microsoft Build 2024 event, Altman even made hints about it.
90-Days Window
Ninety-Day Period”We anticipate the resulting systems to bring us to the next level of capabilities on our path to AGI,” the company’s blog states, referring to the recent training of OpenAI’s next frontier model. This is most likely a clear allusion to GPT-5. In a recent interview, Satya Nadella tacitly acknowledged the same.
Over the next ninety days, the Safety and Security Committee’s primary responsibility will be to assess and improve OpenAI’s procedures and security measures. Following that, the committee will present its recommendations to the entire Board for consideration before making them public.
It’s interesting to note that the GPT-5 release was rumored to occur in the summer, within the 90-day review window. But it’s also assumed that GPT-5 might be revealed after this November’s US elections, which are scheduled for that month.
According to reports, the forthcoming GPT-5 model will have more features and may potentially function as an AI agent. According to reports, it has multimodal capabilities that enable text, images, audio, and video right away. Sora, the newest model of text-to-video generation, will also be included in the system.
Given that the safety alignment team was recently dissolved following the departures of Ilya and Jan Leike, the company’s announcement regarding the Safety and Security committee is probably a preventative measure, even as it prepares for the release of its next model. In addition, a lot of users expressed criticism of the announcement in their fast reactions to it.
