David Robinson Leaves OpenAI, Criticizing AI Safety Practices

David Robinson, a former OpenAI safety researcher, resigned during the week of Oct. 3 and criticized AI companies for moving too quickly as their systems become more capable.

“The companies building this technology aren’t being nearly careful enough,” Robinson wrote in an essay published Oct. 3, 2026. He intends to work outside OpenAI on incentives for safer AI development, though he has not settled on a specific role.

During three and a half years at OpenAI, Robinson says he directed work on the current Preparedness Framework and oversaw safety reports covering 12 frontier launches. He also described OpenAI as standing by its safety practices and maintaining that they are careful enough.

OpenAI’s framework tracks advanced capabilities and links safeguards to capability categories. Its April 15, 2025, update says systems at the “High” threshold need safeguards that sufficiently minimize severe-harm risks before deployment. Systems at the “Critical” threshold also require safeguards during development.

OpenAI leadership makes the final decisions after its Safety Advisory Group evaluates safeguards and recommends actions. Robinson argues that written procedures alone are insufficient. He has called for safety expertise from fields such as aviation and nuclear power, along with more research into whether capable models make safe choices outside testing.