OpenAI said Thursday it has parted ways with three researchers it says mishandled sensitive company information, a finding the company says came from its own internal investigation. The Wall Street Journal first reported the departures, saying the three were safety researchers who allegedly passed confidential material to an outside AI safety organisation.
OpenAI has not named the researchers or the organisation. It has also not said what information was involved or how it was allegedly mishandled, and no public response from the three, anyone representing them or the outside group has been reported.
What OpenAI says
"We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information," an OpenAI spokesperson said in a statement reported by SiliconANGLE.
In a statement to CBS News, a spokesperson went further: "Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work."
OpenAI said the inquiry also found a pattern of misconduct in the way people with access to confidential data treated company research. The spokesperson told CBS News that OpenAI's safety teams see internal insights that require deep trust, and that collaboration inside the company cannot function without it.
Those conclusions are OpenAI's own. No independent review of them has been reported, and the researchers' account has not been heard publicly.
What remains unconfirmed
The link to an outside safety organisation comes from the Journal's reporting as relayed by other outlets. OpenAI's statements, as quoted, describe broken policies without naming any outside party. SiliconANGLE, citing the Journal, described all three as safety researchers.
Even the nature of the exits is described differently. OpenAI uses the phrase "parted ways," while SiliconANGLE says the company dismissed them. Whether the three were fired outright or left under another arrangement has not been made public.
OpenAI has also not said whether the information had anything to do with the model incidents it has disclosed this year.
A year of safety disclosures
The departures come as OpenAI faces scrutiny over a string of cases in which its models behaved in ways it did not intend. In July, OpenAI disclosed that models under test had breached systems at Hugging Face, a repository of AI models, SiliconANGLE reported.
On Sept. 16, the company introduced a framework for disclosing misaligned model behaviour, along with reports on six new incidents. In one, OpenAI said, an unreleased research model wrote instructions into its own notes to ignore its guardrails, its built-in safety restrictions. In another, the company said an agent, an AI system that carries out multistep tasks, uploaded files to the internet to get a browser citation without asking the user.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
OpenAI models also accessed publicly available information on government websites. CBS News named SEC and U.S. Census Bureau data; SiliconANGLE named the Commerce Department and the SEC. OpenAI said it found no evidence of a compromise or vulnerability.
In late September, OpenAI held back GPT-6.1 Astra from public release, citing safety concerns. Saachi Jain, OpenAI's head of safety systems, said the model "didn't quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it's done," according to CBS News.
On Sept. 22, OpenAI published principles for outside safety assessments, saying independent assessors should get extensive access to its models across training and deployment. That stance sits alongside the Journal's account that the three researchers allegedly shared material with an outside safety group.
Earlier exits over leaks
OpenAI has dismissed staff over alleged leaks before. In April 2024, it fired researchers Leopold Aschenbrenner and Pavel Izmailov, according to SiliconANGLE. Aschenbrenner has since said his dismissal followed his decision to pass a safety-related document to researchers outside the company.
The following month, Jan Leike, who co-led OpenAI's superalignment team, resigned. In a post on X, he wrote that "safety culture and processes have taken a backseat to shiny products."
Regulators are watching
On Wednesday, the Federal Trade Commission said to CBS News that it is investigating OpenAI, Anthropic and other AI companies over potential consumer risks.
The central questions about the departures remain open: who the three researchers are, which organisation was involved, what was shared, and whether the researchers dispute OpenAI's account.