Skip to content
NewsLeadership

OpenAI Safety Employee Resigns Citing 'Broken' Culture Days After Firing Three Researchers

· by Pondero Newsdesk

The short version

OpenAI parted ways with three safety researchers on October 1, 2026 over leaked infrastructure details, then a fourth safety staffer quit on October 3 saying the company's culture is broken.

OpenAI Safety Employee Resigns Citing 'Broken' Culture Days After Firing Three Researchers

Two separate exits from OpenAI's safety ranks landed within 48 hours this week. Three researchers lost their jobs over how they handled confidential company data, and days later a fourth, longer-tenured safety staffer quit on his own terms, publicly disputing the company's internal culture.

What

OpenAI parted ways with three members of its safety team, Jasmine Wang, Tomek Korbak, and Mikita Balesni, after they shared confidential company information with an outside AI-safety organization, the Wall Street Journal reported October 1, 2026, per The Hacker News. An OpenAI spokesperson said the company's investigation "confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work." Citing Bloomberg, the same report said the leaked material concerned OpenAI's infrastructure architecture; the outside organization that received it was not named.

Two days later, David Robinson, who said he led the writing of safety reports for OpenAI's major product launches over three and a half years, resigned and published an essay in The Atlantic saying the company's "culture is broken," per TechCrunch. Robinson argued OpenAI's "iterative deployment" approach to shipping products guarantees periodic failures as systems grow more capable, and said frontier labs need the kind of redundancy found at nuclear plants or busy airports. OpenAI spokesperson Drew Pusateri responded that the company is "making significant changes to strengthen security in our research and testing environments" and expanding work with third-party evaluators.

Two different dissent tracks open on the same safety team

A leak-driven firing and a values-driven resignation are not the same event, and neither alone proves an organizational pattern. But an AI-tool operator weighing OpenAI's vendor reliability now has two live threads to track rather than one: a confidentiality breach serious enough to cost three researchers their jobs, and a public resignation letter arguing the company lacks the operational discipline of safety-critical industries. Robinson's essay echoed a September 9 resignation by former OpenAI and Anthropic researcher Jacob Coxon, who said the two companies were "gambling with our lives," per TechCrunch, a comment that prompted Anthropic CEO Dario Amodei to outline a plan to pace frontier development. Neither OpenAI statement addressed whether the two incidents are connected.

What to watch next

A fifth safety-aligned departure within the next week would turn this from a two-event cluster into a trend worth a dedicated follow-up. OpenAI has not disclosed which outside organization received the leaked infrastructure details, nor whether the three dismissed researchers plan any public response of their own.

Sources