OpenAI Safety Leader Quits, Saying Outside Pressure May Matter More
OpenAI Safety Leader Quits, Saying Outside Pressure May Matter More
David Robinson spent three and a half years at OpenAI, helping lead safety transparency and overseeing the company’s preparedness framework and model-safety reports. His departure last week landed amid mounting concern over safety staffing, reported model misbehavior and a series of security incidents involving advanced AI systems.
In his Atlantic essay, Robinson argued that the problem was not a lack of good intentions but an organisation trapped in constant motion. “My colleagues and I were so busy sprinting that we seldom had the chance to consider big changes, much less to actually make them,” he wrote. That is why, he said, “stronger incentives for safety — coming from outside the company — are a big part of getting this right.”
His warning follows a broader run of public departures from major AI labs. Critics of the industry’s pace say frontier companies need less of Silicon Valley’s habitual confidence and more of the redundancy, planning and humility expected of nuclear plants or busy airports — systems designed so a routine human error cannot become a catastrophe.
OpenAI rejects the suggestion that it is simply racing ahead. A company spokesperson said it is strengthening safety and security as capabilities rise, including through third-party evaluations, improved monitoring and more responsible model training. “We’re making sure our models don’t become more capable than we can safely manage and secure,” the spokesperson said, adding that OpenAI will “pause training or hold back models” when it needs to slow down.
Robinson has not set out a fixed next step, but his direction is clear: work from outside the lab to make the risks more legible — and to increase the cost of ignoring them. Whether external scrutiny can reshape the incentives of the AI race is now the sharper question.
Write a comment