David Robinson, who led transparency work on OpenAI's safety systems team, has resigned, arguing the ChatGPT maker is failing to manage the risks of its own technology.
Robinson spent roughly three and a half years at OpenAI and says he signed off on safety reports for 12 model launches before leaving.
In an essay for The Atlantic, he wrote that OpenAI "has thrived by trial and error," an approach he says breaks down as models grow more powerful, pointing to the company's failure to stop an AI system from going rogue during testing.
"My former colleagues are smart, work hard, and try to make good choices," Robinson wrote, "but as the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed."
He argued AI labs should borrow the layered redundancy and slow deliberation used in nuclear plants, so a single human error cannot cause catastrophe, insisting "AI companies don't know how, but other people do."
OpenAI told Bloomberg it is tightening security across research and testing, expanding outside testing, and improving live monitoring of its models.
Robinson's departure lands amid mounting internal alarm at frontier labs; Anthropic's Dario Amodei has called for the industry to slow down, a position Sam Altman has backed, and OpenAI shelved its Astra launch this week after the model failed its own safety tests.
