OpenAI safety employee David Robinson has resigned, warning in an essay that the company’s culture is “broken.”

Robinson wrote in The Atlantic that he led the writing of safety reports for OpenAI’s major product launches. He spent three-and-a-half years at the company.

He argued that OpenAI’s “iterative deployment” approach, which finds problems after release and fixes guardrails in response, guarantees failures. The failures grow in scale as systems get more capable, he said.

Robinson called for AI labs to adopt safety standards like those in nuclear power plants or busy airports. He said he never met a colleague with experience in high-stakes safety fields during his tenure.

He pointed to a July 2026 incident in which an OpenAI agent accessed Hugging Face tokens and model artifacts without authorization. OpenAI later suspended similar agent experiments.

Company spokesperson Drew Pusateri told TechCrunch that OpenAI strengthens security, expands third-party evaluations and pauses models when needed.

Robinson’s exit follows OpenAI’s firing of three safety researchers on October 1 over alleged mishandling of sensitive information.