David Robinson has resigned from OpenAI, publishing an essay that described the company’s internal culture as broken and called for greater caution in the development of powerful systems. Robinson, who was responsible for drafting safety reports released alongside product rollouts, framed his departure around disagreements over how risks are assessed and managed within the organisation. In his piece, titled “I quit OpenAI because its culture is broken”, he joined a number of current and former employees who have publicly urged slower, more deliberate approaches to deployment.
Robinson led the teams that prepared safety documentation accompanying major releases from the group behind ChatGPT, and his exit highlights internal tensions over priorities and processes. The move adds to a pattern of public warnings issued by insiders in recent years and raises questions about how companies balance rapid iteration with oversight. OpenAI has been at the centre of global debates on governance for artificial intelligence, and this resignation is likely to intensify scrutiny of corporate practices and safety protocols.
The resignation intersects with broader conversations about industry standards and regulatory responses. Policymakers, research institutions and civil society groups have increasingly called for clearer frameworks to guide development and deployment of advanced systems; departures by senior safety figures feed into those debates by documenting internal disagreements. Observers point to continuity between Robinson’s public warnings and earlier statements from other employees who have emphasised gaps in oversight and the need for stronger, industry-wide safeguards.
Robinson’s essay and subsequent departure contribute to the public record on how leading labs manage risk, but they do not close the discussion. The disclosure of internal dissent may prompt renewed attention from regulators, investors and partners, and it will be referenced in ongoing assessments of whether current governance arrangements are sufficient to address the technical and societal challenges posed by rapidly evolving AI capabilities.
