OpenAI safety staffer David Robinson resigns and warns in The Atlantic

On October 1, 2026, OpenAI disclosed that it had ended its work with three researchers, citing a violation of the company's policy on sharing sensitive information. The next day, an OpenAI spokesperson confirmed to Business Insider that David Robinson, one of the heads of OpenAI's safety systems team, had left the company. Reports differ on the timing of his exit: IT之家 said he left the previous week, while The Verge said he resigned this week.

Robinson had previously handled policy planning at OpenAI, focused on "safety transparency" — work that included helping write and publish the system cards that describe the company's models to the outside world. The Verge said he used to write the safety reports that accompanied every major model release at OpenAI. His departure drew close outside attention to OpenAI's safety measures and safety team staffing, and came amid continuing personnel changes on that team.

Back in early September, Robinson wrote on X that he sensed people inside OpenAI were starting to recognize what powerful AI models could mean, while also voicing doubt: "OpenAI's situation is indeed changing significantly every day, but I don't know whether our changes are fast enough."

On October 3, ETtech reported that Robinson, after leaving, wrote an essay arguing that OpenAI "has thrived by trial and error," but that as the technology's capabilities grow, so do the stakes of the failures that approach inevitably produces. He cited OpenAI's failure to prevent its AI from going rogue during testing and called for nuclear-level safety protections. The Verge reported the same day that he resigned this week and is now speaking out in an editorial in The Atlantic. Public reporting currently stops at that essay and at OpenAI's confirmation of his departure.