Why OpenAI Dismissed Three Safety Researchers—and What Remains Unclear

OpenAI says it parted ways with three employees after an internal investigation found they mishandled sensitive company information outside established procedures. The Wall Street Journal reported that the matter involved sharing information with an outside AI-safety organization. Those are separate points: the company has described its policy finding, while the reported account adds context about the alleged sharing. 112

What OpenAI says happened

An OpenAI spokesperson said the three violated company policies for accessing and handling sensitive information. The spokesperson said the company’s investigation found that they handled it outside established procedures and broke the trust required for their work. That is OpenAI’s stated rationale; the reporting cited here does not reveal the underlying evidence behind the finding. 12

The Wall Street Journal’s account, as reported by TechCrunch, says the researchers allegedly shared confidential information with an outside AI-safety organization. AFP reported that the work involved an external organization that evaluates AI models. 111

What has not been disclosed

OpenAI has not publicly confirmed the researchers’ identities or identified the outside organization. The company also has not specified what information was involved. 1112

That leaves important questions unanswered: what kind of material was shared, what the recipients did with it, and whether the information concerned a safety issue, model evaluation, or another subject. The available reporting does not establish the answers—or show that the alleged sharing caused a specific harm. 1112

How this compares with the 2024 firings

Coverage has drawn a comparison with OpenAI’s 2024 firings of Leopold Aschenbrenner and Pavel Izmailov over alleged information leaks. The cases share a broad issue—how the company handles confidential information—but the reporting cited here does not establish that the conduct, circumstances, or motives were the same. 9

The unresolved question about raising safety concerns

The dismissals put a difficult question in view: how can employees bring genuine safety concerns to outside experts or authorities without improperly disclosing protected company information? The available evidence does not show that these researchers were punished for whistleblowing, nor does it establish that their alleged disclosures posed a particular security risk. 1112

For now, the clearest account is limited: OpenAI says its investigation found policy violations, and reporting says an outside AI-safety organization was involved. The identities, information shared, and broader circumstances remain unconfirmed publicly.