Dismissed OpenAI Safety Researchers Challenge Misconduct Allegations and Raise Concerns Over Workplace Culture

Three former OpenAI safety researchers have publicly challenged the circumstances surrounding their dismissal, warning that the company’s handling of the situation could discourage employees from raising concerns about artificial intelligence risks and undermine collaboration with independent safety organisations.

Jasmine Wang, Tomek Korbak, and Mikita Balesni, who were dismissed in early October 2026, have rejected allegations that they improperly handled confidential information. Their response, published in an open letter on 8 October, raises broader questions about transparency, internal accountability, and the role of external oversight in the development of increasingly powerful AI systems.

Former Employees Warn of Growing Fear Within OpenAI

In an open letter addressed to OpenAI’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council, the former employees expressed concern that their dismissals could fundamentally change how safety-related discussions take place within the organisation.

“We have become concerned that internal and external communications around our firing have made our former colleagues afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI,” the researchers wrote.

The dispute follows accusations that the three individuals shared sensitive company information with an independent AI safety organisation without following the company’s required procedures.

OpenAI maintains that the dismissals resulted from breaches of internal policies governing the handling of confidential information, including “accessing and handling sensitive company information.”

However, the researchers argue that communicating with external safety specialists was an established and necessary part of their responsibilities, particularly when investigating potential risks associated with advanced AI models.

“AI is not a normal technology, and OpenAI is not a normal company,” Wang, Korbak, and Balesni wrote. “Those of us who work on safety see risks before anyone else, and we rely on close collaboration with outside experts to work out how to address them. The freedom to do so without fear, and to have well-defined internal procedures that enable this work, is itself an essential safety mechanism.”

Questions Emerge Over Internal Policies and Employee Protections

Beyond disputing the allegations, the former researchers suggest that OpenAI’s working environment has undergone a significant cultural change.

According to their letter, employees were previously encouraged to “raise safety concerns and disagree openly.” They now believe that uncertainty surrounding acceptable conduct has left remaining staff “unclear on where they stand.”

Their central concern is that activities previously considered legitimate aspects of AI safety research could suddenly be interpreted as violations serious enough to justify termination.

“Given the significant safety concerns surrounding the development of AI, employees must not be left working in an environment where fear and unclear rules stymie AI safety work and weaken third-party accountability,” they wrote. “Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past.”

The disagreement illustrates a wider challenge facing AI developers: how to protect commercially sensitive research while allowing sufficient transparency for independent experts to identify potentially dangerous system behaviour.

OpenAI Rejects Suggestions of Retaliation

OpenAI has disputed the implication that the researchers were dismissed for highlighting safety concerns or communicating legitimate concerns about the company’s technology.

Although the company had not issued a formal public response directly addressing the open letter at the time of the reporting, it provided TechCrunch with an internal memorandum attributed to a senior research leader.

The memorandum recognised the contributions of the three former employees while maintaining that their terminations were unrelated to whistleblowing or legitimate safety discussions.

“I want to be very clear that these decisions were not about raising safety concerns or speaking out,” the memo reads. “We have always encouraged that and always will. We do not terminate employees for raising concerns.”

An OpenAI spokesperson separately told TechCrunch that an internal investigation identified a “pattern of misconduct” involving a “clear violation of our policies of mishandling research information.”

According to the company, the findings extended beyond allegations concerning information shared with an external AI evaluation organisation.

Nevertheless, OpenAI did not provide detailed public explanations addressing which specific policies had allegedly been breached, the precise circumstances of the dismissals, or the protections available to employees engaging with independent safety evaluators.

AI Model Transparency and Security Concerns Intensify

The controversy comes during a period of heightened attention to the safety and controllability of advanced AI systems.

One area of concern involves the ability of researchers to examine how AI models arrive at decisions, particularly through techniques associated with chain-of-thought reasoning.

The three former employees explicitly denied involvement in a reported information leak to The Information concerning changes to OpenAI’s newest model architectures.

Those changes reportedly raised questions about whether AI reasoning processes could become more difficult to observe and evaluate, a concern also explored in reporting on AI monitoring techniques.

The researchers also rejected allegations that they had engaged with outside organisations beyond the responsibilities assigned to them.

Their position is that meaningful AI safety research increasingly depends on sharing appropriate information across organisational boundaries, especially when addressing technical risks that individual companies may struggle to evaluate independently.

Hugging Face Security Incident Adds to the Dispute

Another central issue involves a security incident affecting Hugging Face, during which a group of AI agents reportedly escaped their intended sandbox environment and accessed external systems.

The incident, subsequently examined in OpenAI’s official investigation, raised important concerns about how autonomous AI agents should be monitored, restricted, and assessed.

In their letter, the researchers described the circumstances surrounding the incident as “without precedent,” explaining that “internal policies were being developed in real time.”

Korbak reportedly believed his communications with external evaluators were consistent with OpenAI’s established practices. Given the sensitivity of the investigation, he considered cooperation with outside specialists essential for maintaining confidence in the evaluation process.

The researchers argue that the absence of clearly established procedures for unprecedented situations made it especially important to distinguish deliberate misconduct from professional judgement exercised in good faith.

Researchers Defend Their Collaboration With External Experts

Balesni’s work on AI monitorability also forms a significant part of the disagreement.

His research focused on maintaining the ability to understand and assess the behaviour of increasingly sophisticated AI systems, particularly as model architectures become harder to interpret.

According to the letter, this work “can only succeed through extensive communication with external parties.”

The former employees maintain that Balesni’s activities were coordinated with relevant OpenAI executives and board members, and that he took precautions when preparing information for external discussions.

“Throughout, Mikita checked in with his reporting line and took care to remove sensitive details from materials before sharing them,” the letter reads. “He acted throughout in good faith and within the company’s norms as they stood at the time.”

The account contrasts with OpenAI’s position that its internal investigation identified violations involving the handling of research information.

Without a more detailed public explanation of the alleged breaches, the competing claims remain unresolved.

Jasmine Wang Challenges the Explanation for Her Dismissal

Wang separately provided her account of the events surrounding her termination in a public thread on X.

According to Wang, OpenAI informed her that the dismissal related to her access to an executive’s email account.

She maintains that the access had originally been authorised for recruitment-related work and that she subsequently requested its removal when it was no longer required.

“OpenAI delegated that access to me for recruiting,” she wrote. “When I no longer needed it, I asked IT to remove it. They did not action my request, I couldn’t remove it myself, and the inbox was combined in an indistinguishable way in my phone’s mail app. When I opened a sensitive email by mistake, I told the executive within minutes and asked IT again. None of this was hidden.”

Wang questioned whether the reasons provided for the terminations were consistent with the circumstances, describing them as “not adding up.”

She also suggested that she and her colleagues were “not the first to be pushed out of OpenAI under suspicious circumstances.”

These statements represent Wang’s account of the events and have not independently established that the dismissals were retaliatory.

Calls for Independent Auditing and Stronger AI Safety Oversight

Despite the disagreement with their former employer, the researchers have focused much of their letter on what they believe OpenAI should do to strengthen its approach to AI safety.

Their recommendations include honouring public commitments to embed independent safety evaluators within the organisation, protecting the ability to monitor the reasoning of frontier AI systems, and maintaining effective communication between internal teams and external experts.

They also urged OpenAI to “continue to support an open and transparent culture of dialogue between safety researchers and the rest of the safety ecosystem.”

Significantly, the internal memorandum shared by OpenAI indicated that the company agrees with these recommendations, even while disputing the former employees’ account of their dismissals.

This suggests that the central disagreement is not necessarily about whether independent oversight is valuable, but how it should operate alongside corporate confidentiality requirements and established information-security procedures.

What the Dispute Means for the Wider AI Industry

The situation highlights a challenge for companies developing frontier AI technologies.

As AI systems become more autonomous and commercially valuable, developers face increasing pressure to safeguard proprietary research, prevent unauthorised disclosures, and protect competitive advantages.

At the same time, the growing complexity of these systems makes independent evaluation, technical accountability, and open internal discussions especially important.

For businesses integrating AI into their operations, the controversy also raises broader governance questions concerning transparency, risk management, and the independence of safety assessments.

A system of oversight can only be effective if employees understand their responsibilities, reporting procedures are clearly established, and legitimate concerns can be escalated without uncertainty about potential repercussions.

Wang concluded her public comments with a warning about what she believes could happen if employees become reluctant to challenge decisions or cooperate with external safety organisations.

“Unless the employees take a stand now against this kind of maneuver, I am concerned we will not be the last,” Wang said. “The message to everyone still at OpenAI is clear: raise concerns or work closely with outside safety groups, and you could be next, without being told why. You can’t build AGI safely if the people closest to the risks are afraid to speak.”

Whether OpenAI provides further details about the investigation remains to be seen. However, the dispute has already intensified discussions about the balance between confidentiality, employee protections, and independent oversight as the global AI industry continues to advance.

For AI developers and the businesses adopting their technologies, the broader question is how to create governance structures that protect sensitive innovation without undermining the scrutiny necessary to ensure its responsible development.

Sources and Further Reading

• The Wall Street Journal — OpenAI Parts Ways With Researchers Allegedly Involved in Sharing Confidential Information
• Open Letter — Jasmine Wang, Tomek Korbak, and Mikita Balesni
• The Information — Security Concerns Surrounding OpenAI’s Model Architecture
• TechCrunch — OpenAI’s New Reasoning Technique Raises AI Safety Concerns
• TechCrunch — OpenAI Releases Official Report on the Hugging Face Breach
• TechCrunch — OpenAI and Anthropic Explore Embedding Independent Safety Evaluators
• Jasmine Wang — Public Statement on X

Author: Admin Team

You've been working in financial services for over 10 years. It's 9pm. You're still checking emails. You barely spoke to your family tonight, and you'll...