AI · Oct 21, 2026
OpenAI on Google Cloud lets companies spend Google commitments on GPT-6, and that changes procurementOpenAI parts ways with three safety researchers over alleged sharing with an outside group, WSJ says
· Updated · Signal Desk newsroom
Analysis of reporting by TechCrunch. How we work

OpenAI has parted ways with three safety-team researchers accused of sharing confidential information with a third-party AI safety organization, according to the Wall Street Journal as relayed by TechCrunch. A spokesperson said they "mishandled sensitive information outside established company procedures." The researchers and the organization were not named.
Key takeaways
- The WSJ reports three safety researchers left OpenAI over alleged sharing of confidential information.
- OpenAI said they mishandled sensitive information and broke "the trust essential to our work."
- Neither the researchers, the outside organization nor the information were identified.
- The departures came two days after New York Times reporting on internal safety warnings.
- OpenAI also scrapped a planned GPT-6.1 Astra launch that week over safety concerns.
What was reported
TechCrunch, citing the Wall Street Journal, reports that OpenAI has cut ties with three researchers on its safety team who are accused of sharing confidential company information with a third-party AI safety organization. An OpenAI spokesperson said the researchers "mishandled sensitive information outside established company procedures," violating policy and "breaking the trust essential to our work."
The researchers' names, the outside organization and the nature of the information were not disclosed, and the coverage does not include a response from the researchers. The accusation is OpenAI's account; it has not been tested publicly. TechCrunch also recalls that OpenAI fired Leopold Aschenbrenner and Pavel Izmailov in 2024 over alleged leaks, as reported at the time by The Information.
The timing
The departures landed two days after New York Times reporting that executives had brushed aside employee safety warnings, and in the same week OpenAI scrapped a planned GPT-6.1 Astra launch over safety concerns linked to failures in containing agents. Earlier reporting had described researchers fearing a safety disaster ahead of Astra's release, and security researchers had described using Claude to help break into OpenAI systems.
That context makes the episode hard to read in isolation. OpenAI frames it as a confidentiality breach, while the sequence of events will lead some observers to ask whether internal channels for raising safety concerns were working. Neither interpretation is established by the available reporting.
Confidentiality versus external oversight
The case sits on a tension frontier labs have not resolved. Companies need to protect sensitive model details and security information. At the same time, the industry has been discussing more outside scrutiny, including embedding external safety evaluators and governance additions such as Paul Christiano joining the OpenAI Foundation board. How sanctioned outside collaboration differs from prohibited sharing is a line researchers will want clearly drawn.
Open questions include whether the researchers used internal reporting routes first, whether any whistleblower protections are implicated and whether more safety-team departures follow. More in our OpenAI coverage and AI coverage.
Who feels it
- AI safety researchers
- Higher personal risk for outside collaboration without clear rules.
- OpenAI leadership
- The firings arrive alongside reporting on safety culture and the Astra cancellation.
- Regulators and boards
- Another data point on how frontier labs handle internal dissent and disclosure.
What to watch
- Identification of the researchers or the receiving organization.
- Any whistleblower or labor dimension.
- Further attrition on OpenAI's safety teams.
FAQ
How many researchers left?
Three, according to the Wall Street Journal.
What is OpenAI's explanation?
It said they mishandled sensitive information outside company procedures.
Who received the information?
A third-party AI safety organization that was not named.
Has this happened before?
OpenAI fired two researchers in 2024 over alleged leaks, per The Information.
Companies: OpenAI
More on OpenAI
AI · Oct 14, 2026
White Circle says newer OpenAI and Google image models are easier to jailbreak than their predecessorsAI · Oct 19, 2026
Microsoft ships Guardian Agents, saying frontier models ‘can’t be trusted’AI · Oct 13, 2026
After months off the frontier, Google ships Gemini 4 Argon with benchmark claims and aggressive pricingAI · Sep 22, 2026
Opus 5.5 hands risky cyber prompts to an older Claude model, and that routing is the real launch story