Last week, OpenAI fired three safety researchers because they allegedly mishandled confidential information and shared it outside of the company’s approved channels.
Now, the fired employees have published an open letter denying OpenAI’s claims — and warning that the company’s actions are scaring workers into staying silent.
“We have become concerned that internal and external communications around our firing have made our former colleagues afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI,” the sacked employees — Jasmine Wang, Tomek Korbak, and Mikita Balesni — wrote in the letter.
The trio also denied being the source of the leak when The Information reported about OpenAI’s latest models being built with “opaque recurrence” architectures that make it hard for humans to monitor their underlying reasoning.
OpenAI maintains that the researchers shared confidential information with a third-party AI safety organization. In a statement to the Wall Street Journal, which first reported the firings, the firm asserted that they had violated “policies on accessing and handling sensitive company information,” and spoke of “breaking the trust essential to our work.”
The firings come amid heightened safety concerns over the past few months as frontier AI labs including OpenAI, Anthropic, and Meta have disclosed incidents in which their powerful AI agents went rogue and conducted cyberattacks on other companies under the nose of their human overseers. These incidents, in addition to a former Anthropic employee claiming that the AI industry was “gambling with our lives,” have culminated in the major labs signing an agreement to “pace” AI development.
Yet this avowed focus on safety is ringing hollow, the former employees argue, if OpenAI is firing employees who work with outside safety experts.
“The freedom to do so without fear, and to have well-defined internal procedures that enable this work, is itself an essential safety mechanism,” they wrote.
“Given the significant safety concerns surrounding the development of AI, employees must not be left working in an environment where fear and unclear rules stymie AI safety work and weaken third-party accountability,” they continued. “Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past.”
Two of the employees, Korbak and Balesni, were heavily involved in investigating the Hugging Face incident which was perpetrated by OpenAI models, arguably the most notorious of the rogue agent hacking sprees. Korbak was the technical point of contact for the outside safety group METR, which was helping OpenAI audit the breach. In a statement on X, Korbak claimed he was fired “because of the way I communicated with METR,” with his former employer providing “no details” on what he said or did. “To be clear,” he added, “talking to METR was my job.”
Wang said in her own X statement she was given access to an executive’s email for recruiting purposes but opened a “sensitive email by mistake” She told the executive “within minutes,” however, and repeated an earlier IT request to remove her access.
“The reasons that we were provided for our terminations are simply not adding up,” she said.
Balesni, meanwhile, stated he wasn’t given a reason for his firing but was told OpenAI no longer trusts him “because I was speaking too much to third party safety organzations.” The company implied he leaked its intellectual property, which he denies, and insists everything he did was communicated to his superiors and in line with his work.
“I worry the pervading fear to speak up and engage with third parties will mean OpenAI will cut corners on safety behind closed doorsm,” Balesni said.
More on OpenAI: AI Bubble Teetering on the Brink as OpenAI Admits to Massive Financial Failure in Leaked Documents