OpenAI safety lead resigns, criticizes company culture in Atlantic essay

David Robinson, who authored OpenAI’s safety reports for every major model release, resigned this week and publicly criticized the company’s safety culture in an essay published by The Atlantic. Robinson described Silicon Valley’s approach to AI development as driven by “extreme confidence” and “perpetual sprints,” with “unimpeded optimism” that underestimates potential risks. He argued that the industry’s ‘move-fast-and-break-things’ ethos is fundamentally incompatible with the risks posed by frontier AI systems.

Robinson called for the adoption of “nuclear-level safeguards,” comparing the oversight needed for AI development to that of nuclear power plants or busy airports. He emphasized the need for layers of redundancy and careful planning to prevent human error from leading to catastrophic outcomes. His critique extends beyond procedural changes, framing the issue as a cultural one within the tech industry.

The former safety lead is the latest in a series of high-profile departures from prominent AI labs by researchers and safety workers who have gone on to publicly warn about AI risks. Robinson’s essay follows similar public statements from former employees at Anthropic, Google DeepMind, and other labs, who have raised alarms about the potential dangers of advanced AI systems.