David Robinson, a security leader at OpenAI, has left the company and publicly criticized its culture, calling it "broken." Robinson argues that the pressure to release new products continuously prevents the company from adopting the level of care necessary for safe development.
Robinson's core concern centers on the speed-at-all-costs approach: the company's optimism that it can solve security problems as they arise will lead to increasing security failures as AI systems become more capable. This represents a fundamental disagreement over risk management philosophy—the belief that safety can be addressed retroactively versus the need for proactive safeguards built into development.
OpenAI responded to these claims by asserting that it does pause training and retain models when necessary to ensure secure development of its capabilities. The dispute highlights a growing tension in the AI industry between rapid innovation cycles and the infrastructure required to manage emerging risks at scale.