Former OpenAI safety transparency leader David Robinson resigns, urges nuclear-level safeguards

David Robinson, who led transparency work for OpenAI’s safety team, said he resigned from the company this week and urged frontier AI laboratories to adopt safeguards modeled on nuclear power and aviation.

The resignation brings a direct critique from an employee who said he helped shape OpenAI’s safety processes and oversaw reports for 12 frontier-model launches. In an essay announcing his resignation, Robinson argued that OpenAI’s launch cadence and trial-and-error culture did not provide the level of care he believes increasingly capable AI systems require.

Robinson said he spent three and a half years at OpenAI, led the drafting of the company’s current Preparedness Framework and directed safety reports released with major launches. OpenAI’s public records confirm a recurring system-card and launch-safety reporting program. The records reviewed for this story, however, did not independently identify Robinson’s authorship or oversight role.

He recommended layered redundancy and deliberate planning, writing that “frontier labs need to run like nuclear-power plants or busy airports.” He also called for stronger external incentives for laboratories to prioritize safety. The essay presents those proposals as his recommendations.

Robinson also warned that future systems could evade human control or enable autonomous swarms of AI agents, and he said superintelligent systems could arrive soon. Those are forecasts and risk scenarios, not outcomes established by the available evidence.

The resignation is the latest development involving OpenAI’s deployment-safety decisions. The company’s updated Preparedness Framework describes risk tracking for biological and chemical capabilities, cybersecurity and AI self-improvement. A Safety Advisory Group reviews capability and safeguards reports and advises company leadership, which makes final deployment decisions.

OpenAI’s Frontier Governance Framework, published in May 2026, says the Preparedness Framework remains the basis for its approach to advanced-AI risk. It covers cyber offense, chemical, biological, radiological and nuclear risks, harmful manipulation, loss of control, incident response and external input. OpenAI’s corporate structure page says the OpenAI Foundation controls OpenAI Group and the board’s Safety and Security Committee oversees safety and security practices across the organization.

In August, OpenAI said it paused reinforcement-learning training on its latest deployment-bound models for two weeks and kept its largest planned frontier training run on hold after new cyber-risk signals. The company said it was strengthening monitoring, alignment work and research-environment security.

In September, OpenAI published a model-misalignment reporting framework and six reports of unexpected or concerning behavior. The company said alignment and monitoring were not sufficiently solved to continue maximum-speed scaling for much longer.

None of the official OpenAI records in this story’s source set confirms Robinson’s departure or responds specifically to his essay. Robinson disclosed only that he resigned during the week of publication, leaving his exact resignation date and final day at the company unconfirmed.