OpenAI safety employee resigns, saying its “culture is broken”

David Robinson, who says he led the writing of safety reports for major OpenAI product launches, has resigned. In an essay published in The Atlantic, he said the company’s “culture is broken.” Robinson said he had spent three-and-a-half years at OpenAI and was “among the longest-tenured employees” there.

Robinson argued that debate over AI safety should address company culture as well as specific rules and laws. OpenAI’s approach of deploying systems iteratively and improving safeguards in response to problems, he wrote, “guarantees periodic failures,” whose scale grows as systems become more capable. He cited a recent breach of Hugging Face systems by OpenAI agents and continuing reports of rogue agents.

He said frontier AI companies should operate with layers of redundancy and careful planning, like nuclear power plants or busy airports. Robinson added that he had not encountered colleagues with experience in those fields or in keeping the financial system from collapsing.

OpenAI spokesperson Drew Pusateri said the company is improving safety measures, including security in research and testing, responsible model training, third-party evaluation and real-time monitoring. He said OpenAI pauses training or holds back models when needed.

Robinson also called for broader questions about AI alignment, saying current measures of how well systems match human values are “coarse.” He said he chose to speak out, and that stronger external incentives for safety are important. Business Insider first reported his departure.