As OpenAI begins training a new model, it establishes a safety committee

While it starts training its next AI model, OpenAI has established a Safety and Security Committee, which the AI startup announced on Tuesday, which will be chaired by CEO Sam Altman and other board members.

The committee will be led by directors Bret Taylor, Adam D’Angelo, and Nicole Seligman, as OpenAI announced in a corporate blog.

The generative AI chatbots from Microsoft-backed OpenAI, which can converse like humans and generate visuals in response to text prompts, have raised questions about safety as AI models grow more sophisticated.

Ilya Sutskever, the former chief scientist, and Jan Leike, the head of OpenAI’s Superalignment team, which made sure AI stayed in line with the goals, departed the company earlier this month.

Days after the high-profile departures, CNBC reported that OpenAI had dissolved the Superalignment team earlier in May, less than a year after the company established it, with some team members being relocated to other groups.

The new committee will be in charge of advising the board on matters pertaining to security and safety for OpenAI’s operations and projects.

Over the next ninety days, its first assignment will be to assess and refine OpenAI’s current safety procedures. After that, it will present its findings to the board.

OpenAI announced that it will make an update to the public on the recommendations implemented following the board’s study.

Jakub Pachocki, the recently appointed Chief Scientist, and Matt Knight, the head of security, are among the group’s other members.

The business will also confer with other specialists, such as John Carlin, a former Department of Justice official, and Rob Joyce, a former US National Security Agency cybersecurity director.

The new “frontier” model that OpenAI is training is not described in any further detail other than to say that it will advance its systems to the “next level of capabilities on our path to AGI.”