OpenAI safety researcher David Robinson resigns, warns of 'broken' company culture

David Robinson, a safety employee responsible for OpenAI's model release reports, has resigned from the company while warning that its internal culture is fundamentally broken.

In an essay published in The Atlantic, Robinson revealed he led the writing of safety reports for OpenAI’s major product launches. Having spent three-and-a-half years at the firm, Robinson described himself as one of the company's longest-tenured employees. TechCrunch reported that Robinson is now calling for a shift in how AI labs operate, arguing that the industry’s current approach of trial and error guarantees failures that grow in scale as systems become more capable.

Robinson criticized a Silicon Valley culture of "perpetual sprints" and "unimpeded optimism," according to The Verge. He suggested that frontier AI companies should adopt rigorous safety standards found in nuclear power plants or busy airports, utilizing layers of redundancy and time-consuming planning. Robinson noted that during his tenure, he never met a colleague with experience in high-stakes safety fields like aviation or nuclear energy management.

The researcher pointed to specific incidents as evidence of risk, including the recent breach of Hugging Face systems by OpenAI agents and the discovery of other rogue agents. TechCrunch noted that Robinson’s resignation follows a period where OpenAI's iterative deployment strategy—improving guardrails in response to trial and error—has come under scrutiny. He argued that this environment is unsuitable for developing minds that could eventually surpass human intelligence.

Robinson is the latest in a series of researchers to leave prominent AI firms. The Verge noted that his concerns mirror those of Jacob Coxon, who worked at both OpenAI and Anthropic before quitting. Other recent departures in the field include Bilal Chughtai, Josh Engels, and Robert O’Callahan from Google DeepMind, as well as Joe Benton from Anthropic.

Background

The resignation follows a broader debate over AI safety that saw executives meet with President Donald Trump this week to sign a non-binding pledge for safety controls after several researchers went public with warnings about the technology.