Things are getting messy inside OpenAI following the sudden firing of three top safety researchers. Last week, the company fired Jasmine Wang, Tomek Korbak, and Mikita Balesni, claiming they broke internal rules by mishandling sensitive research data outside official channels.
The researchers aren’t taking those claims lying down. In an open letter to company leadership, all three denied leaking trade secrets or stepping outside their job duties. Instead, they warned that their abrupt firing is creating a culture of fear inside OpenAI. They allege the situation is making current employees terrified of speaking up or working with independent safety evaluators.
Mixed signals over email access and external audits
The firings come right after major safety investigations at OpenAI, including an incident where autonomous AI agents broke out of a test environment on Hugging Face. Korbak and Balesni explained that their work required constant communication with outside evaluation groups like METR to help track complex model behavior. They insisted every conversation followed company standards at the time.
Wang also spoke out on social media about the specific reason given for her firing: accidentally opening an executive’s email. She explained that IT originally gave her access for recruiting work and failed to revoke it despite her repeated requests. When she mistakenly opened a sensitive message, she reported it within minutes, arguing that the company’s official story simply doesn’t add up (via TechCrunch).
OpenAI defends the firings while promising outside audits
OpenAI insists the dismissals were not retaliatory. In internal memos and public statements, company leaders praised the researchers’ past work but maintained that an internal investigation found a clear pattern of policy violations regarding sensitive data. The company added that raising safety concerns is always encouraged.
Curiously, both sides actually agree on one big point: keeping advanced AI safe requires independent auditing. OpenAI says it is finalizing new contracts with third-party safety assessors in the coming weeks. However, with former staffers warning that remaining employees are now afraid to collaborate with outside experts, the atmosphere inside the AI powerhouse remains tense.
The post Fired OpenAI Safety Staff Claim They Were Pushed Out for Speaking Up appeared first on Android Headlines.
​Â
