OpenAI Parts Ways With Three Safety Researchers After Probe

OpenAI has parted ways with three members of its safety team after an internal investigation found they mishandled sensitive company information, according to the company’s statement reported by TechCrunch. The Wall Street Journal reported that the information was allegedly shared with an outside AI-safety organization. Neither the organization nor the material has been identified.

The distinction matters: OpenAI described a policy violation, while the reported external sharing remains an allegation with few public details. The company said the conduct violated its rules for accessing and handling sensitive information and occurred outside established procedures. The available reporting does not describe what information was involved or provide the researchers’ account.

What the company has and has not disclosed

OpenAI said its investigation confirmed that the three people mishandled sensitive information and violated policies. That is the company’s characterization of its own investigation, not a public account of the underlying evidence. The report does not establish whether the information was shared intentionally, what safeguards or procedures were involved, or what the outside organization received.

Names circulated on social media, but TechCrunch said it had not confirmed them. The researchers and the outside organization therefore remain unidentified in the supplied reporting. That restraint is important in a personnel matter where an unverified name can turn an allegation into an apparent fact.

The unresolved issue with the clearest bearing on safety oversight is whether the researchers had raised concerns internally before the alleged sharing. TechCrunch reported that this was unclear. OpenAI has said it has internal channels for safety concerns, but the existence of those channels does not show whether these employees used them or how any report would have been handled.

The dispute sits in a wider debate about safety reporting

The departures came two days after a New York Times report, as described by TechCrunch, said OpenAI employees had warned about the company’s safety practices. OpenAI told the Times it takes security concerns seriously, has internal reporting channels and recognized a need to move faster. Those separate reports provide context, but they do not establish that the three researchers were dismissed for raising safety concerns.

TechCrunch also placed the departures amid reported security incidents involving OpenAI agents and the company’s decision to scrap a planned GPT-6.1 Astra launch over safety concerns. The article noted another reported case: OpenAI fired researchers Leopold Aschenbrenner and Pavel Izmailov in 2024 over alleged leaks. These events do not establish a pattern in the present case, but they make clear why the boundary between protecting confidential information and raising safety concerns deserves careful scrutiny.

For AI companies, enforcing access rules and maintaining credible internal safety reporting are both consequential. This case establishes that OpenAI says its policies were violated; without detail about the information, process or internal reporting, it cannot answer whether the controls worked fairly or whether an internal safety concern was part of the story. TechCrunch’s report says that last point remains unresolved.