OpenAI has fired three researchers from its security and alignment teams following an internal investigation that the company says confirmed violations regarding the handling of sensitive information. The case is significant because it directly affects the areas responsible for monitoring the risks of its own artificial intelligence systems, and because it coincides with a time when several of those same employees had publicly expressed doubts about the company's direction.
The information was published by The Wall Street Journal and subsequently reported by the-decoder.com. So far, OpenAI has not publicly confirmed the names of those fired, nor has it detailed what information was leaked or which organization received it.
Who are the fired researchers?
According to the U.S. newspaper, the three individuals affected are Jasmine Wang, Tomek Korbak, and Mikita Balesni. Korbak was part of OpenAI's security team, while Wang and Balesni worked in the alignment area, which is responsible for ensuring that AI systems behave in accordance with the goals intended by their developers.
A company spokesperson stated that an internal investigation confirmed violations of OpenAI's policies regarding the handling of sensitive information. However, it was not specified what those violations consisted of or where the information was allegedly sent.
What OpenAI confirms and what remains unclear
The company's statement is limited to noting that the investigation detected violations of its internal policies. It does not specify the documents involved or the external organization that allegedly received the information.
The Wall Street Journal itself does not identify that organization either. The newspaper describes it generically as an external group dedicated to artificial intelligence safety and presents the leak as an attribution, not as a fact verified with public evidence.
Therefore, it is important to separate what is confirmed from what is not. It is a fact that three researchers left the company and that OpenAI alleges a violation of its internal rules. It has not been established, however, what the information contained or who ultimately received it.
Korbak’s role as a liaison with METR and Redwood Research
One detail in the report provides relevant context. Tomek Korbak served as the technical point of contact between OpenAI and two external organizations, METR and Redwood Research. Both had been tasked with analyzing how certain AI agents developed by OpenAI managed to bypass security controls and access third-party systems, including Hugging Face.
In other words, part of Korbak's duties consisted precisely of exchanging technical information with external security evaluators. The Wall Street Journal does not establish any direct link between that work and his subsequent firing, so attributing causality would be premature. But the data helps to understand how delicate it can be to distinguish between authorized collaboration and a leak when the position requires constant communication with third parties.
The departure of a fourth researcher
According to an anonymous account on the social network X, a fourth security researcher, David Robinson, reportedly left OpenAI shortly after the three firings. This is an independently unverified source, and the company has not made any official comment regarding this departure, so it should be treated with caution.
Previous statements on AI risks
What is publicly known is that all four had made statements in September regarding the dangers of advanced artificial intelligence. Korbak wrote that he was not in agreement with many of OpenAI's decisions. Balesni placed the probability of AI eventually causing the extinction of humanity at more than 10%.
Wang, for her part, had signed a public petition in favor of slowing down the development of artificial intelligence systems. Robinson supported the idea that the race to create an AI capable of improving itself could be considered reckless.
What is the relationship between those criticisms and the firings?
That the four researchers had expressed public reservations about the industry is a documented fact. That those opinions were the cause of their departure, however, is not stated by any of the sources consulted by The Wall Street Journal.
OpenAI attributes the firings to a violation of its rules on sensitive information, not to the statements made by those involved. These are two distinct behaviors—publicly opining on risks and sharing internal documents without authorization—with different employment consequences, and the report provides no evidence that the two were conflated in this case.
It is also unknown whether the three fired individuals share the version offered by the company. None of them have responded publicly according to the information available so far.
An episode with pieces still to be clarified
The confirmed balance of the episode is concrete: three firings indirectly acknowledged by OpenAI through an internal investigation, and a fourth departure attributed to an anonymous source. All of them affect personnel linked to the company's security and alignment areas.
The exact content of the leaked information, the identity of the recipient organization, and the position of the fired researchers themselves remain to be known. Until those details are clarified, what is verifiable is the departure of four professionals who had spoken publicly about the risks of artificial intelligence, and OpenAI's official explanation, which focuses on an alleged breach of its confidentiality rules.
Frequently Asked Questions
Who are the researchers fired by OpenAI?
According to The Wall Street Journal, those fired are Jasmine Wang, Tomek Korbak, and Mikita Balesni. Korbak worked on the security team, while Wang and Balesni were part of the alignment area.
What information was leaked and to whom?
It has not been publicly confirmed. OpenAI has not detailed which documents were involved or which external organization might have received them.
What was Tomek Korbak’s relationship with METR and Redwood Research?
Korbak was OpenAI's technical contact with both organizations, which were tasked with investigating how certain AI agents from the company bypassed security controls and accessed external systems like Hugging Face.
Who is David Robinson and why is he mentioned alongside those fired?
David Robinson is a fourth security researcher who, according to an anonymous account on X, left OpenAI shortly after the three firings. OpenAI has not confirmed or commented on this departure.
Are the firings related to the researchers’ public criticisms?
There is no evidence of this. OpenAI attributes the firings to violations of its rules on sensitive information, not to the opinions previously expressed by those involved regarding the risks of AI.
Has OpenAI confirmed the names of those fired?
No. The information comes from The Wall Street Journal. OpenAI has only confirmed, through a spokesperson, that an internal investigation detected violations of its sensitive information rules.
