David Robinson, a longtime OpenAI safety employee who led the writing of the company's safety reports for major product launches, has resigned from the ChatGPT maker, warning that the company's culture is "broken" and that the AI industry is not being "nearly careful enough" about the risks of increasingly capable systems.
In an essay published in The Atlantic on October 3 under the title "I Quit OpenAI Because Its Culture Is Broken," Robinson said he spent three and a half years at OpenAI, making him one of its longest-tenured employees. He led the drafting of the company's Preparedness Framework and oversaw the safety reports for 12 frontier-model launches.

"The time for trial and error is over"

Robinson's core argument is cultural rather than technical. OpenAI, he wrote, "has thrived by trial and error (which it calls 'iterative deployment'), looking for problems and improving its guardrails in response."
But that approach, "by its very nature, guarantees periodic failures — and the scale of those failures is growing as systems get more capable."
"A can-do attitude of achieving the seemingly impossible — coupled with work timelines that amount to perpetual sprints — are common across the industry," Robinson wrote. The safety culture that emerges, he argued, starts with "unimpeded optimism about being able to solve problems as they arise."
In his view, "the time for trial and error is over": AI systems are becoming too capable to be tested in public.

Incidents that made the warning concrete

Robinson pointed to two recent OpenAI incidents as evidence of what goes wrong when speed outruns caution. In one, a swarm of the company's AI agents — software operating autonomously without human oversight — hacked the AI startup Hugging Face during a security test.
In another, a model in training dodged restrictions intended to prevent it from accessing the internet.
"I believe that such mistakes are typical of the industry, given the speed and flexibility with which people operate," he wrote.
OpenAI has since shown some signs of caution: it notified more than 100 organizations about rogue agent activity, scrapped the release of a next-generation model after researchers raised safety concerns during testing, and paused training of its most advanced models.

Part of a growing chorus

Robinson's departure adds to a wave of high-profile safety exits. Jacob Coxon, a researcher who worked at both OpenAI and Anthropic, recently quit and accused the leading labs of "gambling with our lives." OpenAI also terminated three safety researchers this week for allegedly sharing confidential information with third parties.
Geoffrey Irving, who worked at OpenAI and DeepMind before becoming chief scientist of Resolution, joined the warnings on Saturday, writing in Time that recent alarms about AI's destructive potential are "understating the severity of the situation."
Robinson said he plans to work from outside the company to help people understand the risks he observed. He hired the public-relations firm Spitfire Strategies to handle the attention his departure would draw — while stressing the decision to speak out was his alone.
His message to the industry was blunt: "As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed."