OpenAI dismisses three safety researchers over information sharing
On 1 October 2026 The Wall Street Journal reported that OpenAI had dismissed three researchers from its safety team for allegedly sharing confidential company information with a third-party AI-safety organisation. The three were identified as Jasmine Wang, Tomek Korbak and Mikita Balesni. An OpenAI spokesperson said the company had 'parted ways with three individuals for violating our policies on accessing and handling sensitive company information', and that an investigation had confirmed the individuals mishandled sensitive information outside established company procedures, breaking trust essential to the company's work. Neither the recipient organisation nor the specific material was named publicly; Bloomberg reported that at least some of the information under investigation related to the architecture of OpenAI's infrastructure. The dismissals came in the same week that OpenAI cancelled the GPT-6.1 Astra release over failed safety evaluations and received an investigative subpoena from the California attorney general.
Why It Mattered
The episode sits at the unresolved centre of frontier-lab governance: how safety staff can raise concerns outside their employer when the subject of those concerns is the employer. OpenAI has a specific history here, from the 2024 non-disparagement controversy to the departures that followed the dissolution of its superalignment team, and this is the first instance of the company dismissing named safety researchers for external information sharing. The facts are contested and incomplete — OpenAI frames it as unauthorised access to infrastructure details, which is a security matter rather than a whistleblowing one, while the involvement of an outside AI-safety organisation invites the opposite reading. That ambiguity is itself the historically interesting part, because no mechanism currently exists to resolve it. There is no statutory whistleblower protection in the United States covering disclosure of AI safety information, no designated regulator to receive such disclosures, and no agreed line between a company's legitimate security interest and the public interest in knowing how frontier systems behave. The timing sharpens the point: the dismissals landed days after an industry pledge premised on self-policing and alongside a state subpoena premised on the opposite assumption. Whether later accounts read this as an internal security breach or as a governance failure will depend on information not yet public, but the structural gap it exposes is already clear.
Who Built It
OpenAI
Applications
- AI Safety
- Corporate Governance
- Whistleblower Policy