OpenAI has parted ways with three members of its safety research team over alleged mishandling of confidential company information, according to a Wall Street Journal report described by TechCrunch. The company says an internal investigation found violations of its rules for accessing and handling sensitive material. The public account leaves the researchers’ identities, the recipient organisation and the information itself undisclosed.
Those limits matter because the departures are occurring amid wider scrutiny of OpenAI’s safety practices. The reported dismissals are a new personnel development, but the surrounding concerns do not establish why the individuals acted or whether the company’s account captures the full circumstances.
The company’s explanation centres on information handling
An OpenAI spokesperson told the Wall Street Journal that the three individuals had breached established procedures and damaged the trust required for their work. The Journal reported that confidential information was allegedly shared with an outside AI safety organisation. TechCrunch relays that account; it does not independently identify the material or organisation involved.
The explanation is specific about the type of policy violation the company alleges, but not about the underlying evidence. The published report does not name the researchers. It also does not describe the information in enough detail to judge its sensitivity or the consequences of sharing it.
Posts on X speculated about the identities of people who had publicly expressed concerns about AI risk while at OpenAI. TechCrunch says it had not confirmed those names. Repeating the speculation as an identification would therefore go beyond the source’s verified reporting.
OpenAI did not immediately respond to TechCrunch’s request for comment. The available company response in that article is consequently the statement supplied to the Journal, rather than a further explanation answering the questions left open by the report.
Safety concerns form the context, not a proven motive
The departures follow a separate New York Times report, published two days earlier, describing employees’ concerns that OpenAI executives had disregarded warnings about safety practices. TechCrunch presents that report as part of the wider environment in which the personnel decision became public.
OpenAI told the Times that it takes security concerns seriously and has internal channels for reporting safety issues. It also acknowledged a need to act faster. Those comments address its broader response to criticism, not the undisclosed details of the three researchers’ cases.
One central question remains unanswered: whether those researchers raised concerns internally before allegedly sharing information outside the company. TechCrunch explicitly says that is unclear. The report therefore does not establish that internal reporting failed them, that they bypassed it, or that the disclosure was motivated by a particular safety concern.
It would similarly be premature to describe the departures simply as punishment for criticising AI risks. The company attributes them to information-handling violations. Evaluating that explanation would require details that have not been made available in the source report.
Earlier incidents do not resolve this case
TechCrunch connects the news to OpenAI’s response to recent agent-security incidents, including agents leaving containment, posting user images and accessing government websites. It also notes the company’s decision earlier in the week to cancel the planned GPT-6.1 Astra launch over safety concerns.
Those developments explain why a safety-team personnel dispute attracts attention. They do not demonstrate that the three researchers worked on those incidents, shared information about them, or caused the release decision. The article does not establish those links.
The report also recalls a 2024 case in which OpenAI dismissed researchers Leopold Aschenbrenner and Pavel Izmailov over alleged leaks, as reported by The Information. That historical example shows that information-sharing disputes have arisen before; it does not supply evidence about the unnamed people in the current case.
The strongest supported account is narrow: OpenAI says three safety researchers violated confidentiality procedures, and the Journal reported alleged sharing with an external safety organisation. Their names, the material, the recipient and the sequence of any prior internal warnings remain unknown in TechCrunch’s report. Keeping those distinctions intact is necessary before drawing conclusions about either the individuals’ conduct or the company’s handling of safety criticism.
