Fired OpenAI researchers deny data mishandling — TechCrunch
Three fired OpenAI safety researchers — Jasmin Wang, Tomek Korbak, and Mikita Balesni — published an open letter denying the company’s allegations of improper handling of sensitive information. They said their dismissal is creating an atmosphere of fear at OpenAI and could prevent employees from openly raising artificial intelligence safety concerns.
As TechCrunch reports, OpenAI fired the three employees last week, citing violations of internal policies on access to and handling of sensitive information. According to the outlet, the dismissals were linked to the alleged sharing of confidential information with an outside organization focused on AI safety.
Researchers’ denials
In the letter, Wang, Korbak, and Balesni stressed that safety specialists must cooperate with outside experts in order to identify and mitigate risks. According to them, OpenAI’s culture previously encouraged employees to voice disagreement and raise safety concerns, but staff now do not understand what conduct could become grounds for dismissal.
The letter’s authors also denied involvement in a leak to The Information about the architectures of OpenAI’s latest models, which are harder to control through monitoring of chains of reasoning. They denied interacting with external parties outside the scope of their job duties.
More current news is available on the UA.News Telegram channel Telegram.
OpenAI’s position
OpenAI did not provide a formal response to the open letter, but shared with TechCrunch an internal memo attributed to the head of its research division. The memo noted the three employees’ contributions to AI safety and denied that the decision to dismiss them was a response to the concerns they had expressed. The company also did not directly answer the outlet’s questions about the specific policies allegedly violated and the circumstances of the dismissals.
Wang separately wrote on X that, according to OpenAI’s explanation, she was fired over access to an executive’s email. She claims that she was granted this access for recruiting purposes and, after completing the work, asked the IT department to revoke it. According to her, she accidentally opened a confidential email in a phone app and informed the executive about it shortly afterward.
The researchers called on OpenAI to uphold its public commitments to integrate independent safety auditors into the organization, preserve the ability to monitor frontier models, and support open dialogue between safety researchers and the broader community of specialists.
