Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect
Three OpenAI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—have disputed misconduct claims after being fired last week, publishing an open letter that warns their dismissal signals a troubling shift in the company's culture. They argue that the terminations will have a chilling effect on employees' willingness to raise safety concerns and collaborate with external experts, practices they claim were previously integral to OpenAI's operations.
OpenAI fired the researchers after alleging they mishandled sensitive company information by sharing it with a third-party AI safety organisation. The researchers denied involvement in a leak to The Information about less monitorable architectures in OpenAI's models and denied engaging with external parties beyond their job mandates. OpenAI responded with an internal memo asserting the firings were not retaliatory, though the company has not formally addressed the open letter or clarified which specific policies were violated.
- Three OpenAI safety researchers fired for allegedly sharing confidential information, deny misconduct
- They warn dismissal creates chilling effect on safety culture and external collaboration
- Incident comes amid broader scrutiny over rogue AI agents and model leaks
New here? Start with this
OpenAI fired three safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—accusing them of sharing confidential information with an external AI safety organisation. The dismissal appears linked to a public leak about OpenAI's systems, although the researchers deny they leaked information to journalists and claim they worked within their assigned responsibilities. The dispute concerns whether the researchers violated company policy regarding confidential information.
OpenAI's safety researchers work on identifying and addressing potential risks in the company's AI systems, a role that traditionally involved collaboration with outside experts and internal discussions about the company's direction. The firing has prompted concern that employees may now be reluctant to raise safety concerns or engage with external researchers if they fear dismissal. OpenAI has said the firings were not retaliatory, but has not publicly specified which policies were violated.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Companies reasonably expect to control access to sensitive technical information, particularly in AI where architectural details could present security risks; if the researchers did share proprietary information with external parties without authorisation, dismissal is a proportionate consequence regardless of their intentions. The principle that employees must follow proper channels for raising concerns—rather than unilaterally deciding to share company information with outside organisations—is fundamental to maintaining trust and security when handling sensitive work.
The case against
AI safety is a matter of genuine public interest, and researchers at leading AI companies must be able to collaborate with the broader safety research community and escalate concerns without fear of retaliation; if OpenAI terminates employees for attempting to raise safety issues through external channels, it sends a chilling signal that discourages the critical oversight the field needs. The researchers dispute the core allegations, and OpenAI's failure to publicly substantiate specific policy violations or engage substantively with their concerns raises legitimate questions about whether this was retaliation for safety advocacy.