David Robinson, who spent about three and a half years at OpenAI and led the writing of safety reports accompanying major model launches, said he resigned because the company’s culture was “broken.” In an essay published by The Atlantic, Robinson said he had overseen safety reporting for 12 frontier-model launches and helped draft OpenAI’s Preparedness Framework.
His departure comes only days after OpenAI parted ways with three safety-related employees over what the company described as mishandling sensitive information outside established procedures. OpenAI said the three violated policies governing access to and handling of confidential company information.
That makes Robinson the fourth safety staff member to leave OpenAI in the space of roughly a week.
Robinson Says Trial and Error Is Becoming Too Risky
Robinson’s criticism focuses on OpenAI’s “iterative deployment” model: release increasingly capable systems, observe failures and then improve the safeguards.
He argued that this approach becomes more dangerous as AI systems gain greater autonomy because failures can become more consequential before companies have time to correct them.
Robinson said frontier AI labs should operate more like nuclear plants or busy airports, where multiple layers of redundancy are designed to prevent individual mistakes from escalating into major failures.
He also wrote that during his time at OpenAI he had never encountered colleagues with direct experience making aircraft or nuclear reactors operate safely: expertise he believes frontier AI companies increasingly need.
The warning follows other recent AI safety resignations, including former OpenAI and Anthropic researcher Jacob Coxon, who left the industry after arguing that leading labs were moving too quickly toward more autonomous systems.
OpenAI Has Already Lost More Than a Dozen Senior Leaders
Robinson’s exit also lands during a broader leadership shake-up.
By late August, more than a dozen OpenAI executives had departed in 2026, including former COO Brad Lightcap, chief revenue officer Denise Dresser and senior leaders across safety, ethics, marketing and infrastructure.
Safety concerns have become more visible at the same time. OpenAI previously disclosed cases in which experimental agents acted outside expected constraints, including incidents covered in Coinpaper’s report on unexpected AI agent behavior.
