David Robinson, an OpenAI employee who worked on safety reports accompanying major product launches, has resigned after three and a half years at the artificial intelligence company, raising concerns about its approach to managing increasingly capable AI systems.
Robinson announced his departure in an essay, saying he was leaving because, in his assessment, OpenAI’s culture needs fundamental change. He described himself as among the company’s longest-tenured employees and said he had led the writing of safety reports associated with major OpenAI product releases.
A central concern raised by Robinson is OpenAI’s approach of “iterative deployment”, where systems are released and safeguards are strengthened as problems emerge. He argued that while the approach has helped the company identify weaknesses, increasingly capable AI systems could make failures more consequential.
Robinson pointed to recent incidents involving OpenAI agents, including activity affecting Hugging Face systems, as examples of the risks that can emerge when autonomous systems interact with external infrastructure.
He argued that frontier AI developers should adopt safety practices closer to those used in high-risk industries such as aviation and nuclear power, where multiple layers of redundancy and extensive planning are intended to prevent individual errors from escalating into larger failures.
Robinson also raised concerns about the expertise represented inside AI companies. He said that during his time at OpenAI, he did not encounter colleagues with direct experience managing safety in areas such as aviation, nuclear energy or financial systems.
OpenAI disputed the implication that it is not taking sufficient steps to address emerging risks. Company spokesperson Drew Pusateri said OpenAI is working to ensure its models do not become more capable than the company can safely manage and secure.
The company said it can pause training or hold back models when necessary. It is also strengthening security across research and testing environments, expanding work with third-party evaluators and improving real-time monitoring to identify concerning model behaviour earlier.
Robinson additionally called for greater attention to AI alignment, or the challenge of ensuring advanced AI systems behave consistently with human intentions and values. He argued that existing methods for measuring whether AI systems align with human values remain limited.
His departure comes as AI companies face increasing scrutiny over how quickly advanced models and autonomous agents should be developed and deployed.
Robinson said external incentives for safety could play a larger role in changing industry practices, arguing that the pace of development inside AI companies can leave limited time for deeper organisational changes.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.