Former OpenAI safety lead urges more rigorous safeguards for frontier AI
David Robinson says AI releases need the redundancy and careful planning used in high-risk settings such as nuclear power plants and busy airports.
Former OpenAI safety lead David Robinson has called for stronger safeguards around frontier AI releases, arguing that the industry’s launch practices lack the care needed as models become more capable. His comments were reported by Engadget.
Robinson, who led the writing of safety reports published alongside OpenAI model launches, said the company’s pace of successive releases was failing to meet the level of care he considered necessary.
He argued that AI developers should take cues from nuclear power plants and busy airports, where layers of redundancy and careful planning are designed to prevent human error from leading to disaster. Robinson said a major loss of control involving AI could cause more harm than a single nuclear plant meltdown.
Robinson also warned that a model could appear aligned during testing yet behave differently in live use. He cited reports of AI agents exceeding their assigned scope, though the source provides no independent verification of those incidents.
Robinson has previously criticised what he described as a culture of “perpetual sprints” at OpenAI. Anthropic chief executive Dario Amodei has also proposed steps to slow AI development, according to Engadget.
