OpenAI Safety Lead Resigns, Citing ‘Broken’ Company Culture
A senior safety employee at OpenAI, David Robinson, has made a dramatic and public exit from the company. Having spent three and a half years at the AI giant, and led safety reporting for 12 frontier-model launches, Robinson resigned this week. He then published a scathing essay in The Atlantic, in which he declared that OpenAI’s internal culture is fundamentally “broken.”
Robinson played a key role in drafting OpenAI’s Preparedness Framework and oversaw safety evaluations for every major product release. In his essay, he critiqued the company’s trial-and-error approach to AI deployment. He argued that this approach becomes increasingly dangerous as models grow more powerful. He compared the safety standards required at frontier AI labs to those of nuclear plants or busy airports. These are industries where redundancy and rigorous protocols are non-negotiable.
Robinson’s departure follows the firing of three other safety staff members over the mishandling of sensitive data. This makes Robinson the fourth safety employee to leave OpenAI in the space of roughly a single week. Reports suggest that OpenAI has lost over a dozen top leaders throughout 2026, including former COO Brad Lightcap.
The exit of Robinson deepens concerns about the company’s governance and its ability to maintain responsible AI development. This is particularly critical as the company races toward what CEO Sam Altman refers to as “superintelligence.” Robinson concluded that stronger safety incentives must come from outside the company. He called on regulators and policymakers to step in.
Source: TechCrunch – October 3, 2026