OpenAI safety leader David Robinson resigns and criticizes company culture
Robinson says OpenAI's pace and culture fall short of the care needed for increasingly capable systems; the company says it is strengthening safeguards.
A departure followed by a public warning
David Robinson, who worked on safety transparency and system cards at OpenAI, has resigned, Business Insider reported. The publication said it first reported his departure on Friday and that he left the previous week. In an essay published Saturday and covered by Business Insider and TechCrunch, Robinson argued that OpenAI’s culture and rapid launch cycle do not allow enough care for increasingly capable AI systems. That is his assessment of the company’s practices, not an independently established finding that a particular model is unsafe. His departure is separate from OpenAI’s reported parting with three other safety researchers over alleged policy violations.
TechCrunch reported Robinson’s concern that a trial-and-error approach can produce failures as systems become more powerful. He cited recent agent security incidents as reasons to change incentives and planning. OpenAI spokesperson Drew Pusateri told TechCrunch that the company is strengthening security in research and testing, expanding third-party evaluations and improving real-time monitoring. The company also said it would pause training or hold back models when needed. Those are OpenAI’s stated measures and intentions; the reviewed reports do not independently measure their effectiveness or establish whether Robinson’s requested cultural changes will follow.