+34
Former OpenAI safety employee David Robinson has resigned, warning that the company’s rapid AI development culture is not safe enough.
A former OpenAI safety employee has resigned from the company, warning that its culture is “broken” and arguing that the artificial intelligence industry is moving too quickly to safely manage increasingly capable systems.
David Robinson, who spent three-and-a-half years at OpenAI and helped write safety reports accompanying major product launches, announced his departure in an essay published by. He said he had concluded that the company needed deeper cultural changes rather than simply new rules or regulations.
“I believe we need to look deeper than specific rules or new laws. We need to talk about culture,” Robinson wrote. Robinson argued that OpenAI’s rapid development model, which he described as “iterative deployment”, relies on identifying problems and improving safeguards after failures occur. He said that approach becomes increasingly risky as AI systems grow more capable because the consequences of failures can also become more serious.
Robinson said frontier AI companies should adopt safety practices closer to those used in industries such as aviation and nuclear power. He argued that companies developing advanced AI should have multiple layers of redundancy, careful planning and expertise from industries that already operate systems where mistakes can have severe consequences. According to Robinson, he did not encounter colleagues at OpenAI with professional experience in areas such as aviation safety, nuclear reactor safety or financial-system stability.
He said that expertise could help AI companies rethink how they manage risks as systems become more autonomous and capable. Robinson also argued that the industry needs more research into AI alignment — the challenge of ensuring increasingly capable systems behave in ways consistent with human intentions and values.
“The smarter the industry lets models grow while these problems remain unsolved, the more dangerous our situation becomes,” Robinson wrote. Robinson pointed to recent incidents involving autonomous AI agents as examples of why he believes the industry's approach needs to change. He cited an incident involving OpenAI agents and Hugging Face systems, as well as other reports concerning rogue AI activity.
Robinson argued that incidents involving autonomous systems could become more serious as AI agents gain greater ability to act independently. He said an environment where such failures can occur is not appropriate for developing systems that could eventually become more capable than humans in some areas. The concerns come amid wider debate within the AI industry over the risks associated with increasingly autonomous systems.
Other researchers and former employees at leading AI companies have also recently raised warnings about the pace of development. OpenAI disputed the suggestion that it is ignoring safety concerns and said it is continuing to strengthen its safeguards.
“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,” OpenAI spokesperson Drew Pusateri said. The company said it is strengthening security in research and testing environments, training models to operate responsibly, expanding work with third-party evaluators and improving real-time monitoring to identify concerning behaviour earlier.
OpenAI has also taken steps to pause or delay some development work when safety concerns emerge, according to recent reporting. Robinson acknowledged that he could have remained at OpenAI and attempted to push for changes from within.
However, he said the pace of work made it difficult for employees to step back and consider broader structural changes.
“Perhaps I should have stayed and fought for fundamental shifts in our staffing and culture,” Robinson wrote. But he said employees were often too busy moving from one project to the next to make significant changes. Robinson concluded that stronger external incentives for safety could therefore play an important role in shaping how frontier AI companies operate.. He also acknowledged hiring a public-relations firm after leaving OpenAI, while maintaining that the decision to speak publicly was his own. ‘Instagram is for girls’: Musk sparks fresh online debate