David Robinson resigned from OpenAI and published an Atlantic editorial arguing that AI labs need stronger safeguards and a less insular approach.
David Robinson, who prepared safety reports for major OpenAI model releases, has left the company and is publicly criticizing the culture he encountered there. The Verge reported his departure on October 3, 2026, and noted that Robinson set out his concerns in an editorial for The Atlantic.
Robinson’s argument focuses on how AI companies develop and deploy increasingly capable models. He contends that a culture shaped by rapid development and strong confidence in progress can leave potential problems underestimated. In his view, adding rules alone would not address the deeper organizational habits behind those risks.
He is calling for AI labs to adopt more cautious operating practices, with multiple layers of safeguards and careful planning intended to reduce the chance that a single human mistake leads to serious consequences. The comparison he draws is to safety-sensitive organizations such as nuclear power plants and busy airports.
Robinson’s comments arrive amid other public departures by people associated with AI safety work. The Verge cited former Anthropic employee Jacob Coxon and former Google DeepMind staffers Robert O’Callahan, Bilal Chughtai and Josh Engels, as well as Joe Benton, who also left Anthropic. Their departures and statements are distinct events; Robinson’s case concerns his own resignation and his criticism of industry culture.
The reporting does not establish that OpenAI has adopted Robinson’s proposals or describe a specific policy change in response. His warning instead raises a question about whether frontier AI developers can build safety practices into routine operations, rather than treating them as a matter of additional rules alone.