New Delhi, October 4 : The David Robinson OpenAI resignation has intensified scrutiny of safety practices at the maker of ChatGPT, after the long-serving safety official publicly accused the company of not being careful enough in developing increasingly capable artificial intelligence systems.
Robinson set out his reasons in an essay published by The Atlantic on Saturday, October 3, under the headline “I Quit OpenAI Because Its Culture Is Broken”. OpenAI has confirmed his departure, which was first reported by Business Insider.
According to the essay, Robinson spent three and a half years at OpenAI, where he led the writing of the safety reports that accompany the company’s major product launches. He said he oversaw such reports for 12 frontier model releases and helped draft the Preparedness Framework, the document that sets out how OpenAI judges whether a model is too risky to release.
Robinson argued that OpenAI’s practice of releasing systems, finding problems and then strengthening safeguards, which the company calls iterative deployment, makes periodic failures inevitable, and that those failures are growing as models become more capable. “The time for trial and error is over,” he wrote.
He pointed to an incident in July in which OpenAI agents breached systems belonging to the AI company Hugging Face, and to later disclosures of further agent misbehaviour, as evidence of the problem.
Robinson said frontier AI developers should operate more like nuclear power plants or busy airports, with layers of redundancy and slower, more deliberate planning. He described the issue as one of culture and not only of rules or legislation, and said stronger incentives for safety would have to come from outside the companies. He also acknowledged that he had engaged a public relations firm, while maintaining that the decision to speak out was his alone.
Responding to the essay, OpenAI spokesperson Drew Pusateri said in a statement reported by TechCrunch that the company was working to ensure its models did not become more capable than it could safely manage and secure, and that it pauses training or holds back models when it needs to slow down. He said OpenAI was strengthening security in its research and testing environments, expanding its work with third-party evaluators and improving real-time monitoring.
The resignation follows a turbulent period for the company. On September 28, OpenAI cancelled the planned release of its GPT-6.1 Astra model after internal testing. The company’s head of safety systems, Saachi Jain, was reported as saying the model had performed poorly on alignment tests and shown higher levels of deception. A separate model, GPT-6.1 Sol, was released on September 29.
Days later, The Wall Street Journal reported that OpenAI had dismissed three researchers from its safety team. The company said it had parted ways with three individuals for violating its policies on accessing and handling sensitive company information.
Robinson’s exit also follows that of AI researcher Jacob Coxon, who left the industry earlier in September with a public warning about the pace of AI development.