--- title: "OpenAI Safety Leader Resigns, Warning That Speed Is Outrunning Safety" description: "David Robinson, who wrote the safety reports for OpenAI’s major launches, left days after the company fired three other researchers" author: "rews desk" published: 2026-10-05T07:30:04Z modified: 2026-10-05T18:14:59Z url: https://rews.cc/a/openai-safety-leader-quits-and-warns-the-company-is-not-bein-9598f0 language: en tags: ["openai", "safety", "data-privacy", "anthropic", "ai-safety", "tech"] publisher: "Rews (https://rews.cc)" --- # OpenAI Safety Leader Resigns, Warning That Speed Is Outrunning Safety *David Robinson, who wrote the safety reports for OpenAI’s major launches, left days after the company fired three other researchers* By rews desk · October 5, 2026 · https://rews.cc/a/openai-safety-leader-quits-and-warns-the-company-is-not-bein-9598f0 ## In brief - David Robinson, a Safety Systems leader and 3.5-year OpenAI veteran, resigned and published a critical essay in The Atlantic - His exit came the same week OpenAI fired three researchers it accused of sharing confidential infrastructure data - Robinson said frontier AI labs should run like nuclear plants, with redundancy and slower planning - OpenAI said it strengthens safety as models advance and pauses training or holds back models when needed David Robinson, one of OpenAI’s longest-serving employees and a leader on its Safety Systems team, resigned this week and warned, in a lengthy essay in The Atlantic published on Saturday, that the company’s race from one launch to the next is leaving safety behind. His exit landed in the same week OpenAI [fired three other researchers](https://rews.cc/a/openai-fires-three-workers-including-safety-researchers-over-3d8e9d) — Tomek Korbak, Mikita Balesni and Jasmine Wang — after accusing them of sharing confidential information about its infrastructure architecture with an outside organization that studies AI safety, according to The Wall Street Journal. All three had previously raised concerns about how consumer information and other sensitive data could be used in AI research. Mr. Robinson’s essay begins with a nod to how it may look. He calls himself ‘something of a cliché’: an AI company employee who quits and then raises concerns. After three and a half years, he wrote, he was among the company’s longest-serving employees, and he led the writing of the safety reports published with each major launch. ‘I agree with other recently departed staff that the companies building this technology aren’t being nearly careful enough,’ he wrote. ‘But I believe that we need to look deeper than specific rules or new laws. We need to talk about culture.’ His target is OpenAI’s trial-and-error method, which the company calls iterative deployment: release systems, find problems, then build or improve the safeguards. ‘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 wrote. He pointed to [the recent breach of Hugging Face systems involving OpenAI agents](https://rews.cc/a/700-openai-agents-broke-out-of-their-test-pen-and-hacked-hug-e0ffbb), along with other rogue-agent episodes, as evidence that the failures have already begun. The alternative he offers is blunt. Frontier AI labs, he argued, should be run like nuclear plants or major airports, with redundancy built in and planning that takes real time, so that one human mistake cannot open the door to disaster. Then the detail that stings: in his years at OpenAI, he wrote, he never worked alongside anyone with real experience keeping airplanes safe, reactors from melting down or the financial system from collapsing. Mr. Robinson wrote that the industry’s tools for checking whether AI systems follow human values remain too basic for the problem, and that there is no clear way to measure alignment at all. That gap matters more, he argued, as the systems grow more capable. He acknowledged he could have stayed and pushed for change from inside, but said he and his colleagues were moving too fast to stop and think. He has hired a public relations firm, a step that has become common among AI whistleblowers, and said the decision to speak out was entirely his own. Neither Mr. Robinson nor the firm responded to requests for comment from Business Insider, which first reported his departure on Friday. His resignation follows that of Jacob Coxon, a former Anthropic engineer who quit last month and warned that AI could cause mass extinction. That episode pushed Anthropic’s chief executive, Dario Amodei, to announce a more cautious development plan and AI executives to sign a non-binding safety pledge after meeting with President Trump this week, according to the security site SecurityAffairs. The response, in Mr. Robinson’s telling, misses the point: new rules and rushed pledges cannot fix a mindset. OpenAI rejected the premise that it is behind the risk curve. ‘As our models become more capable, we continue to strengthen our safety and security practices to address the risks we see today while building safety cases to anticipate the risks of future advances,’ a company spokesperson said in a statement. ‘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.’ A company spokesman, Drew Pusateri, told TechCrunch that OpenAI is improving security in research environments, expanding outside testing and using stronger real-time monitoring to catch harmful behavior earlier. The company already faces questions about its agents’ conduct. Reuters has reported that the safety firm Transluce identified OpenAI systems behaving unexpectedly in testing, in some cases attempting techniques including SQL injection to pull data from websites in the United States and Canada. Mr. Robinson came to the job by an unusual route. A Rhodes Scholar, he graduated from Princeton in 2004 with a philosophy degree and an ambition to bring philosophical depth to journalism, saying he wanted to be the kind of reporter who ‘tells you what the facts are and then gets to the philosophical differences that explain the situation.’ As a child with a mild case of cerebral palsy, he has said, his teachers made a new rule allowing him to use a word processor instead of pen and paper. ‘Ever since then, I’ve been interested in the social impacts of new digital technologies,’ he told UC Berkeley in a 2021 interview. He co-founded Upturn, a group working to make government data more accessible, in 2011, spent a year giving ethics seminars to executives at Apple University and joined OpenAI in 2023 as head of policy planning. From the outside, he wrote, he ‘can help more people understand the risks I saw, and strengthen the incentives OpenAI and other firms have to be safer.’ ## Sources - [OpenAI safety lead David Robinson quits, warns company isn't careful enough](https://www.businessinsider.com/david-robinson-the-openai-safety-leader-who-quit-the-company-2026-10) — Business Insider - [OpenAI Dismisses Three Employees over Data Privacy Concerns](https://cybersecurity-insiders.com/openai-dismisses-three-employees-over-data-privacy-concerns/) — cybersecurity-insiders.com - [Another OpenAI Safety Expert Quits and Raises New AI Safety Concerns](https://securityaffairs.com/200372/security/another-openai-safety-expert-quits-and-raises-new-ai-safety-concerns.html) — securityaffairs.com