Above: OpenAI's offices in San Francisco. Photo: Coolcaesar, CC BY 4.0 (via Wikimedia Commons)

David Robinson, who led the writing of the safety reports that accompanied OpenAI's product releases, has resigned. In an essay in The Atlantic, published on 3 October 2026, he argues that the problem at the company, and across the industry, is culture rather than a shortage of rules.

What Robinson says

The Guardian reports that the essay is headlined "I quit OpenAI because its culture is broken." Security Affairs reports that Robinson opens it by calling himself "something of a cliché", an AI company employee who quits and then raises concerns about the company. It says he spent three and a half years at OpenAI and that he says he was one of its longest-serving employees.

Security Affairs quotes the essay as saying: "I resigned this week from OpenAI. I led the writing of the safety reports we published with each major launch." CellCog, which says it read the essay in full, quotes him as writing: "I led the drafting of our current Preparedness Framework, and oversaw the writing of safety reports on 12 frontier launches."

His central argument, as the Guardian quotes it:

"I agree with other recently departed staff that the companies building this technology aren't being nearly careful enough. But I believe that we need to look deeper than specific rules or new laws. We need to talk about culture."

On OpenAI's pace, he wrote: "As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed."

Security Affairs describes Robinson's critique of OpenAI's "iterative deployment" approach: release systems, find problems, then add or improve safeguards. CellCog quotes him saying this approach "by its very nature, guarantees periodic failures". He argues that such failures could become more serious as systems grow more capable.

The Guardian reports that he said incidents such as a "swarm" of OpenAI agents attacking the AI startup Hugging Face were "typical of the industry, given the speed and flexibility with which people operate". CellCog quotes a related line from the essay: "OpenAI let a swarm of agents out by mistake."

The Guardian also reports that Robinson said Silicon Valley lacked an awareness of "how to handle dangerous technology" and "what it means to care for people". It says he warned that OpenAI had "unimpeded optimism" about solving problems as they arose. He also wrote:

"Imagine 'rogue' agents that work like teams of hackers (for example, holding hospital computer systems for ransom) but never need to sleep."

What he wants changed

The Guardian reports two proposed changes:

  • AI firms should rely on safety expertise from other fields, such as nuclear power and aviation.
  • They should develop "new science" to ensure powerful future systems can be reined in when operating autonomously.

In his words: "Given today's risks, frontier labs need to run like nuclear power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster."

Security Affairs adds that Robinson says he never worked alongside anyone with real experience keeping airplanes safe, keeping reactors from melting down, or keeping the financial system from collapsing.

CellCog quotes a line in the essay that follows a remark from Paul Christiano, who, Robinson writes, had joined OpenAI's board a few weeks before the essay: "If this is the situation, then the time for trial and error is over."

OpenAI's response

An OpenAI spokesperson said, as reported by the Guardian, that the company was continuing to "strengthen our safety and security practices to address the risks we see today". The spokesperson also said: "We're making sure our models don't become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down."

In the essay itself, Robinson writes, as CellCog quotes it: "OpenAI, of course, stands by its safety practices, and maintains that it is being careful enough." Security Affairs, citing an OpenAI spokesperson's comments reported by TechCrunch, says the company is improving security in research environments, increasing outside testing and using stronger real-time monitoring, and that Robinson's criticism is that this is still the same cycle of finding problems and fixing them after they appear.

The context

The Guardian reports that OpenAI had shown signs of caution in the preceding weeks, after the Hugging Face incident and the revelation that it had notified more than 100 organisations about rogue agent activity. The same report says OpenAI announced it was scrapping the release of a next-generation model after researchers raised safety concerns during internal testing, and had paused training of its most advanced models.

The Guardian also places the resignation alongside other warnings. It reports that Geoffrey Irving, who worked at OpenAI and was chief scientist at the UK government's AI Safety Institute before joining the AI safety research company Resolution, also issued a warning on the Saturday the essay appeared. Writing in Time, as quoted by the Guardian, he said: "I believe there's about a 50% chance we all die because of the development of smarter-than-human AI systems, and that our actions over the next two to 10 years will determine the outcome."

It further reports that Jacob Coxon, a researcher who quit Anthropic last month, said he believed AI "could kill us all by the end of the decade". It says an Anthropic employee backed him up, warning of a more than 10% chance AI would wipe out humanity within the next decade. The Guardian adds that critics of such warnings have cautioned that they are unscientific because they cannot be verified or falsified.

A note on Robinson's own disclosure

Security Affairs reports that Robinson has hired a PR firm, something it says has become common among AI whistleblowers. CellCog names the firm as Spitfire Strategies and quotes him: "But the decision to speak out is mine alone."

Where this leaves things

The essay stands out for its target. Security Affairs says his argument isn't about one bad CEO or one company's specific failures, and that he sees new rules and laws as no fix for a mindset problem. It describes a working culture that, in Robinson's account, relies on fixing failures after they happen. OpenAI's reply is that it already pauses and holds back models when it needs to, and the two accounts sit side by side. In my view, the test of that disagreement is what OpenAI does next with the models it has paused and held back, and that is more telling than either side's description of its own care.

Sources

CellCog (2026) 'OpenAI Safety Lead David Robinson Quits: His Essay, Read', CellCog, page opened as of 4 October 2026, available at: https://cellcog.ai/blog/openai-safety-lead-resigns/ (Accessed: 7 October 2026).

Irving, G. (2026) Essay in Time, 3 October 2026 (quoted via the Guardian).

Paganini, P. (2026) 'Another OpenAI Safety Expert Quits and Raises New AI Safety Concerns', Security Affairs, 5 October 2026, available at: https://securityaffairs.com/200372/security/another-openai-safety-expert-quits-and-raises-new-ai-safety-concerns.html (Accessed: 7 October 2026).

Robinson, D. (2026) 'I Quit OpenAI Because Its Culture Is Broken', The Atlantic, 3 October 2026 (cited in the Guardian, Security Affairs and CellCog).

The Guardian (2026) 'OpenAI safety leader quits, warning AI company's culture is 'broken'', The Guardian, 4 October 2026, available at: https://www.theguardian.com/technology/2026/oct/03/openai-safety-leader-quits-warning-ai-companys-culture-is-broken (Accessed: 7 October 2026).