Skip to content
AI Radar by Wagner Solution
CS
Menu
worth noting Business and investment verified update

Dismissed researchers at OpenAI deny leaking information and challenge the reasons for their dismissal

confirmed by 2 independent sources updated October 24, 2026

Three dismissed researchers at OpenAI deny sharing confidential information and say they received no written clarification of their alleged misconduct. The company maintains that rules were violated and denies any connection between the dismissals and the researchers raising safety concerns.

Tomek Korbak, Mikita Balesni and Jasmine Wang denied sharing confidential information in interviews with The Wall Street Journal. According to their accounts, they received no written clarification of which information they had allegedly mishandled. The earlier report of an information leak thus remains a disputed allegation that the researchers deny.

OpenAI said an internal investigation had confirmed violations of its rules for handling sensitive information. It also denied that the dismissals were related to the safety concerns raised by the researchers. According to the report in The Wall Street Journal, investigators examined suspicions that details about internal model architecture had been shared and found communications in the Signal app with disappearing messages enabled on personal phones. The company said Signal was not an approved communication tool, but using it was not the reason for the dismissals.

The researchers expressed suspicions that their dismissals were related to their advocacy for the ability to monitor intermediate steps in model reasoning; this connection has not been confirmed. In an open letter, they warn that remaining employees are afraid to communicate with external experts and raise safety concerns. They call for clear rules on collaboration with external safety organizations.

What changed

Why it matters

The dispute concerns the conditions under which internal researchers can collaborate with independent safety auditors. According to the dismissed researchers, uncertainty around these rules may limit the sharing of safety concerns. However, their accounts and the company statement differ on the reasons for the dismissals.

What was added since the original report

Verified updates

  1. Correction or contradiction

    The researchers deny sharing confidential information.; According to the researchers, they received no written clarification of which information they had allegedly mishandled.; OpenAI said that using the Signal app was not the reason for the dismissals.

    • The researchers deny sharing confidential information.
    • According to the researchers, they received no written clarification of which information they had allegedly mishandled.
    • OpenAI said that using the Signal app was not the reason for the dismissals.
  2. New verified information

    In an open letter, the dismissed researchers warn of fear among remaining employees.; Tomek Korbak and Mikita Balesni took part in investigating the incident involving the Hugging Face platform.; According to Korbak, OpenAI told him verbally that communication with METR was the reason for his dismissal.; Korbak says he spent months raising internal concerns about the declining ability to monitor the reasoning of AI agents.; OpenAI rejects claims of a loss of the ability to oversee AI agents.

    • In an open letter, the dismissed researchers warn of fear among remaining employees.
    • Tomek Korbak and Mikita Balesni took part in investigating the incident involving the Hugging Face platform.
    • According to Korbak, OpenAI told him verbally that communication with METR was the reason for his dismissal.
    • Korbak says he spent months raising internal concerns about the declining ability to monitor the reasoning of AI agents.
    • OpenAI rejects claims of a loss of the ability to oversee AI agents.
  3. New verified information

    Wall Street Journal identifies those dismissed as Jasmine Wang, Tomek Korbak and Mikita Balesni.; Tomek Korbak worked on the safety team.; Jasmine Wang and Mikita Balesni worked on aligning AI behavior with human intentions.; OpenAI has not confirmed the names of those dismissed.; Korbak was the technical contact at OpenAI for METR and Redwood Research.

    • Wall Street Journal identifies those dismissed as Jasmine Wang, Tomek Korbak and Mikita Balesni.
    • Tomek Korbak worked on the safety team.
    • Jasmine Wang and Mikita Balesni worked on aligning AI behavior with human intentions.
    • OpenAI has not confirmed the names of those dismissed.
    • Korbak was the technical contact at OpenAI for METR and Redwood Research.

Relevant practical impact

What this means

01

For a business

For those managing safety research, the case presents a risk of disrupted collaboration with external experts: according to the dismissed researchers, remaining employees are afraid to communicate outside the company and report safety issues.

People and management
What to decide Check whether company policies clearly define how sensitive information may be shared when working with external safety auditors.
More business impacts →
alignment AI safety Chain-of-Thought METR information protection OpenAI

Check the original

Event sources

confirmed by 2 independent sources · 2 publishers, 2 independent. We count feeds from the same owner only once.

4
TechCrunch AI independent context · first detected OpenAI cuts ties with 3 safety researchers, WSJ reports The Decoder (daily AI news) independent context Three firings and a fourth departure shake up OpenAI's safety team The Decoder (daily AI news) independent context OpenAI's safety crisis keeps getting worse and the company keeps making it worse The Decoder (daily AI news) independent context Three fired OpenAI safety researchers say they were never told what they actually leaked