TechCrunch reported on October 3, 2026 that Robinson described OpenAI’s culture as “broken” in an essay published by The Atlantic.
Robinson said he had spent about three and a half years at OpenAI and was among its longest-tenured employees. He criticized the company’s iterative-deployment approach, arguing that trial-and-error deployment guarantees failures whose scale rises with model capability. He called for redundancy, slower planning, stronger external safety incentives, and further work on alignment.
OpenAI spokesperson Drew Pusateri said the company is strengthening security in research and testing environments, using third-party evaluators, improving real-time monitoring, and can pause training or withhold models when needed.
The report does not independently establish Robinson’s account of OpenAI’s internal culture, the scope of the incidents he cites, or whether the measures described by OpenAI address his concerns.