§ Claim under review · Business
"David Robinson, the OpenAI researcher responsible for overseeing AI safety, quit the company after calling its culture 'broken' in an essay published in The Atlantic."
Verdict
Mostly accurate
Confidence
MediumSummary
This is mostly accurate. David Robinson did resign from OpenAI in early October 2026, OpenAI confirmed his departure to Business Insider, and The Atlantic published an essay by him on October 3 titled "I Quit OpenAI Because Its Culture Is Broken." The word "broken" is his own, and the quotes in the post about nuclear plants, layers of redundancy and the end of trial and error match what news outlets quote from the essay. The one real overstatement is his job. He was one leader on OpenAI's safety systems team who led the writing of the safety reports published with major launches, not the single person in charge of AI safety at the company, a responsibility that reporting places with other executives after a July 2026 reorganization. The order of events is also slightly off: he resigned first, and the essay came out the next day explaining why. The post's mention of three fired researchers matches separate confirmed reporting, but no source links those firings to his resignation as cause and effect. His exact job title at the time he left could not be confirmed from an official OpenAI source.
The readings
key figures from the evidencefrontier launches with safety reports Robinson oversaw
Why this verdict
Evidence
An essay by David Robinson titled "I Quit OpenAI Because Its Culture Is Broken" was published by The Atlantic on 2026-10-03. Multiple named outlets quote it directly. Per TechCrunch, Robinson wrote that he led the writing of the safety reports published with each major OpenAI launch, that at three and a half years he is among the longest-tenured employees at the company, and that the culture is "broken." Bloomberg quotes him writing that OpenAI "has thrived by trial and error" and argues the stakes of the resulting failures are rising with capability. Coverage also quotes his proposal that frontier labs run like nuclear power plants or busy airports, with layers of redundancy and careful, time-consuming planning.
Business Insider reported on 2026-10-02, a day before the essay, that Robinson had resigned the prior week, and an OpenAI spokesperson confirmed the departure. Business Insider described him as a leader on OpenAI's Safety Systems team who worked on safety and transparency including system cards, and who previously led the company's policy planning work.
Separately, reporting from July 2026 based on Wired describes OpenAI merging safety into research: safety teams report to Mia Glaese as VP of Research and Safety, with Saachi Jain as interim head of safety systems after Johannes Heidecke left. Robinson is not identified in any source as holding overall responsibility for AI safety at OpenAI.
On the post's secondary claim: WSJ-sourced reporting, with OpenAI confirming the terminations, says three safety and alignment staff were dismissed over handling of sensitive information, reported shortly before Robinson's exit.
Findings
✓ What's accurate 7
- A person named David Robinson resigned from OpenAI in late September or the first days of October 2026. OpenAI confirmed the departure to Business Insider on 2026-10-02.
- The Atlantic published an essay by him on 2026-10-03 under the title "I Quit OpenAI Because Its Culture Is Broken."
- The word "broken" is his, applied to OpenAI's culture, as quoted by several outlets reading the essay.
- He worked in safety at OpenAI. Business Insider describes him as a leader on the Safety Systems team who worked on safety transparency and system cards and had previously led policy planning.
- He states in the essay that he led the writing of the safety reports published with each major launch, and that at about three and a half years he is among OpenAI's longest-tenured employees.
- The caption's further details are consistent with coverage: the "nuclear power plants or busy airports" and "layers of redundancy" language is quoted from the essay, the "time for trial and error is over" line is quoted by The Atlantic's own editor sharing the piece, and the 12 frontier launches figure appears in that same publisher post and in multiple reports.
- The caption's secondary claim about three fired researchers matches WSJ-sourced reporting that OpenAI dismissed Jasmine Wang, Tomek Korbak and Mikita Balesni over handling of sensitive information shared with an outside safety group, with OpenAI confirming the dismissals.
≈ What's misleading 3
- The claim calls him "the OpenAI researcher responsible for overseeing AI safety," a definite-article phrasing that describes a single person in charge of AI safety at the company. The sourcing supports something narrower: he was one leader on the Safety Systems team who led transparency and safety-reporting work and had earlier led policy planning. Reporting from July 2026 places overall responsibility for OpenAI's safety teams with Mia Glaese as VP of Research and Safety, with Saachi Jain as interim head of safety systems. The gap turns "a senior safety person who wrote the safety reports left" into "OpenAI's safety chief left," which a reasonable reader would take as a bigger event than the record shows.
- Date context mismatch, in a small way on sequencing: the claim says he quit "after calling its culture 'broken' in an essay." The reported order is the reverse. He resigned first, the departure was confirmed by OpenAI on 2026-10-02, and the essay appeared on 2026-10-03. The essay explained a resignation that had already happened rather than preceding it.
- Capability extrapolation, applied to the post's framing rather than to a model: the caption links Robinson's exit to the firing of three researchers with "his exit follows OAI firing researchers." The two events are close in time and both involve safety staff, but no source I found connects them causally, and the firings are described by OpenAI as policy violations over handling sensitive information, not as safety-dissent dismissals.
? What's uncertain 3
- Robinson's exact job title and reporting line at the time of resignation. Outlets vary between "a leader on the Safety Systems team," "led transparency work on the safety team," "safety lead" and, in lower-quality aggregators, "head of Safety Systems." No official OpenAI org page or statement giving his title was located.
- Whether he drafted OpenAI's current Preparedness Framework, as the caption says, or helped draft it. Coverage splits between those two formulations, and I could not check the framework document's own authorship.
- Whether OpenAI issued any direct response to the substance of Robinson's criticism beyond a general spokesperson defense of its safety practices.
Sources
9 of 9 linked to recordsThe Atlantic, David Robinson, "I Quit OpenAI Because Its Culture Is Broken," published 2026-10-03
Business Insider (Stephen Council), reporting OpenAI spokesperson confirmation of the resignation, 2026-10-02, as indexed by Techmeme
Bloomberg (Seth Fiegerman), "OpenAI Safety Employee Quits, Calls for Nuclear-Level Safeguards," 2026-10-03
TechCrunch, "OpenAI safety employee resigns, claiming the company's 'culture is broken'," 2026-10-03
The Verge (Terrence O'Brien), "An OpenAI safety employee has quit and is sounding the alarm," 2026-10-03
Adrienne LaFrance (The Atlantic) post linking the essay and describing Robinson as among OpenAI's longest-tenured employees who oversaw safety reports on 12 frontier launches
Calcalist, "The time for trial and error is over," 2026-10
Wired-sourced coverage of OpenAI's July 2026 safety reorganization (TNW, Gizmodo, PYMNTS): safety teams moved under Mia Glaese, VP of Research and Safety, with Saachi Jain as interim head of safety systems after Johannes Heidecke's exit
WSJ-sourced coverage of OpenAI terminating three safety and alignment staff (Jasmine Wang, Tomek Korbak, Mikita Balesni), with OpenAI confirming the dismissals