David Robinson, a former OpenAI safety employee who led work on safety transparency reporting, resigns from the company and argues that OpenAI’s approach to AI risk is not sufficiently cautious. In an essay published by The Atlantic, Robinson says the company has “thrived by trial and error,” and he calls for changes in how AI labs manage safety.
Robinson’s remarks are framed as an industry-wide warning. Multiple outlets report that he urges AI development to adopt safeguards comparable to those used in highly regulated, high-risk sectors, describing nuclear-plant-like practices and “nuclear-level safeguards.” The Guardian adds that Robinson characterizes OpenAI’s internal culture as “broken,” and he contends that safety work and decision-making do not reflect enough care given the pace of technological progress.
The outlets describe Robinson’s role as involving safety reports tied to ChatGPT product releases and note reporting that he had led transparency efforts for OpenAI’s safety team. While all sources agree on the resignation and the call for stronger safeguards, they differ mainly in emphasis: some focus on nuclear-level comparisons, while others highlight cultural and process failures.
