AI Safety · Satire
OpenAI Announces New AI Safety Board Consisting Entirely of OpenAI Employees
The board, which will provide independent oversight of the company's most powerful models, will meet quarterly and report its findings directly to the CEO.
SAN FRANCISCO — OpenAI has established a new AI Safety and Security Committee, the company announced Tuesday, describing it as an important step toward ensuring that its most capable models are developed responsibly.
The committee consists of four members, all of whom are current OpenAI employees, and will be chaired by the company’s chief technology officer.
“Independence is crucial,” said a company spokesperson. “That’s why we’ve ensured that every member of this committee has a deep, personal financial stake in the company’s success.”
The committee’s first act was to approve the training of a new model, which the company described as “our most capable and most safe model to date, which is also our least safe model to date in absolute terms but our most safe relative to its capabilities.”
External researchers praised the initiative.
“It’s great to see a company taking safety seriously,” said one AI researcher, who asked not to be named because she works for a company with a similar arrangement. “Self-regulation is always the best kind of regulation.”
The committee is expected to meet for the first time next quarter, following the model’s deployment.
