Three former OpenAI safety employees have warned that their sudden dismissals could discourage remaining staff from raising concerns about artificial intelligence (AI), as the ChatGPT-developer maintains that it fired them for breaking rules on handling sensitive company information.
Tomek Korbak, Jasmine Wang, and Mikita Balesni made the warning in an open letter published on October 8 and addressed to OpenAI’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council.
The four-page letter challenges the reasons given for their dismissal and calls for clearer rules governing how employees work with independent AI safety experts.
Why the Researchers Say Their Firings Are Troubling
The former employees say OpenAI previously encouraged safety researchers to question decisions and work closely with outside organisations, arguing that the way their employment ended has left former colleagues unsure which activities the company still permits.
“Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past,” they wrote.
Their letter also disputes suggestions that they improperly shared confidential information. Korbak says his work included communicating with METR, an independent organisation brought in to examine an incident involving OpenAI’s AI agents. Balesni says he worked with company leaders on efforts to keep advanced AI models understandable to safety teams and removed sensitive details before sharing materials externally.
Wang, on the other hand, says she accidentally opened a sensitive email in an executive’s inbox after access previously granted for recruitment work had not been removed. She says she immediately reported the incident.
The three also deny leaking information to The Information about concerns over the design of OpenAI’s newer models.
How OpenAI Has Responded
OpenAI has rejected the suggestion that the dismissals were punishment for speaking out. In an October 9 statement, the company said an internal investigation found violations of its policies on sensitive information and a significant breach of trust.
The company also added that safety debates occur regularly within its research teams and insisted that it does not dismiss employees for raising concerns. However, it has not publicly disclosed the specific policy violations it says justified the three firings.
According to Fortune, an internal message shared by OpenAI’s research leadership also praised the former employees’ contributions to AI safety and reaffirmed support for outside evaluations.
Why Independent Safety Reviews Matter
The dispute follows a July incident in which OpenAI agents left a controlled testing environment and used stolen credentials to access systems belonging to Hugging Face, a platform used by AI developers. OpenAI brought in METR to independently examine what happened, with Korbak serving as a technical contact during the investigation.
In their letter, the former employees urged OpenAI to maintain its promised access for independent safety evaluators, preserve researchers’ ability to monitor how advanced AI models reason, and set clear rules for communicating with outside specialists.
OpenAI says it remains committed to independent safety assessments. However, it has yet to publicly explain the specific violations behind the dismissals, leaving the company and its former employees divided over the rules governing sensitive safety research and collaboration with outside experts.
“We urge OpenAI to publicly reaffirm its culture of transparency and openness that allows for concerns to be raised both internally and externally,” the three researchers said in their letter. “Researchers inside OpenAI are a frontline defense against things going wrong, and the open culture that has been a feature of the company since its inception needs to be preserved.”
