David Robinson, who led transparency work on OpenAI’s Safety Systems team, has resigned, Crypto Briefing reported on 3 October, citing a post on X. He oversaw system cards and public disclosures about risks and incidents — the material OpenAI uses to tell outsiders how it assesses its models.
Key points
- Robinson’s reported departure came in late September or early October 2026.
- His work covered OpenAI’s system cards and disclosures about risks and incidents.
- Former safety systems head Johannes Heidecke left in July, before OpenAI moved its safety teams into research.
- The dismissals of three safety researchers were reported on 1 October.
Robinson’s work on OpenAI’s system cards
System cards accompany model releases and describe capabilities, testing and potential problems. Robinson’s role also covered public accounts of risks and incidents, giving him responsibility for work that people outside OpenAI can read and scrutinise.
His professional profiles showed him working on transparency through mid-September 2026 and changed after that point. The departure was reported to have occurred in late September or early October; OpenAI has not publicly confirmed it, according to Crypto Briefing.
Robinson also wrote Voices in the Code, a book about democratising AI. His exit removes the lead for public-facing safety disclosures during a period in which OpenAI is preparing for a potential public listing.
Heidecke’s July exit reshaped OpenAI safety
Johannes Heidecke, the previous head of safety systems, left OpenAI in July 2026. OpenAI then moved its safety teams into its research division, with the combined group reporting to Mia Glaese, VP of Research and Safety.
That change put Robinson’s reported resignation in a team already reorganised after its previous leader left. The work he led — system cards and public risk and incident disclosures — remained part of how OpenAI communicated its safety assessments outside the company.
Crypto Briefing reported that OpenAI lost at least six senior safety figures over roughly two years. It counted Heidecke’s departure in July as the sixth significant exit in that period.
Three researchers’ dismissals reported on 1 October
Robinson’s reported departure came around the time three other safety researchers were reported to have been dismissed. AI Affairs reported that OpenAI parted ways with the researchers after an internal investigation.
The dismissals were reported on 1 October, according to Crypto Briefing. They allegedly handled sensitive information improperly and passed it to an external AI-safety organisation.
