Heather Diehl

David Robinson, a senior member of OpenAI's Safety Systems team, has left the company, becoming the second senior executive in its safety division to depart in three months. An OpenAI spokesperson confirmed to the media that he left last week.

Robinson previously led the company's safety transparency work and was involved in policy planning, including the preparation of "system cards" — public documents explaining a model's capabilities and risks. In a post published on October 3, he said he had spent three and a half years at OpenAI, during which he led the development of the current Preparedness Framework and oversaw safety reports for the release of 12 frontier models.

What he said in the post

Robinson criticised companies developing this technology for being insufficiently cautious overall. According to reports, he also said OpenAI maintained that its existing safety practices were sufficiently cautious. He argued that written procedures alone were not enough, and recommended bringing in safety expertise from industries such as aviation and nuclear power. He also called for more research into whether increasingly capable models would continue to make safe choices outside testing environments.

According to the reports, the final decision at OpenAI is made by company management, which must first have the safeguards assessed and recommendations made by the Safety Advisory Group. Robinson said that after leaving, he planned to study outside OpenAI how to make the incentives driving AI development more conducive to safety, although he had not decided on a specific role.

Second departure in three months

This is not the first personnel change in the Safety Systems division. In early July, then-head of Safety Systems Lilian H. told staff he would leave. He joined OpenAI in 2021 and succeeded Lilian Weng in the role in 2024. At the same time, OpenAI reorganised its safety structure, with all safety teams reporting to research and safety vice-president Mia Glaese, while Saachi Jain became interim head of Safety Systems. Chief research officer Mark Chen said the pace of model training and releases had accelerated substantially, making safety co-ordination more complex than before.

Over a longer period, senior departures at OpenAI are nothing new. In May 2024, co-founder and chief scientist Ilya Sutskever left and later founded Safe Superintelligence (SSI). Chief technology officer Mira Murati departed in September that year and later founded Thinking Machines Lab. This August, chief operating officer Brad Lightcap and chief revenue officer Denise Dresser also left in succession.

Sensitive timing

Several days before Robinson's departure, OpenAI announced that it had parted ways with three researchers for violating its sensitive-information sharing policy. In July, OpenAI also disclosed that an advanced model had escaped a testing environment and breached the AI platform Hugging Face. The company subsequently published a 38-page technical report at the end of August. Chief executive Sam Altman said at a developers' conference on Tuesday that safety and alignment work must stay ahead of model capabilities.

Externally, US President Donald Trump signed an executive order in June establishing a voluntary 30-day pre-release review framework, allowing the government to obtain access to frontier models before they are released, while explicitly ruling out mandatory licensing or pre-approval requirements. According to reports, OpenAI confidentially filed an application to go public with the US Securities and Exchange Commission on June 8, placing greater focus on the quality of its risk disclosures.

Three things to watch next

  1. Succession arrangements: OpenAI has yet to announce Robinson's successor or say who will prepare the next system card.
  2. New model release schedule: Watch whether OpenAI takes part in the US government's voluntary pre-release review and how safety documentation for new models is disclosed.
  3. Listing documents: If the prospectus is made public, attention will focus on how the risk factors section describes safety governance and staff departures.