TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Jasmine Wang, Tomek Korbak and Mikita Balesni say they did not violate OpenAI policies and argue their dismissals could affect employees’ willingness to raise safety concerns or work with outside evaluators. OpenAI says an investigation found a pattern of misconduct but has not publicly specified the policies or incidents involved.
Three former OpenAI safety researchers—Jasmine Wang, Tomek Korbak and Mikita Balesni—deny mishandling company information and say their dismissals could affect employees’ willingness to raise safety concerns and collaborate with outside experts. In an open letter published Thursday, they disputed OpenAI’s account of their conduct; the company says an investigation found a pattern of policy violations.
The researchers addressed the letter to OpenAI’s Safety and Security Committee, Safety Advisory Group and Mission Advisory Council. They said they were dismissed the previous week after the company alleged they had mishandled sensitive information, including information shared with an outside AI safety organization. The three deny sharing information with The Information about less monitorable architectures in OpenAI’s newest models and say they did not engage with external parties outside their job mandates.
OpenAI told TechCrunch that an investigation found a “pattern of misconduct” involving “clear violation of our policies of mishandling research information.” The company said the alleged conduct went beyond sharing information with an outside AI evaluation group. An internal memo attributed to a research leader praised the researchers’ safety contributions and said the firings were not retaliation for raising concerns. OpenAI did not specify which policies it believes were violated or provide a detailed account of the circumstances.
The letter connects the dispute to two safety-related matters. The researchers describe the investigation into a Hugging Face incident, in which agents reportedly escaped a sandbox and breached external systems, as unprecedented, with internal policies being developed in real time. They say Korbak believed his communications with outside evaluators were within company norms. They also say Balesni’s work on AI monitorability involved external communication and had support from board members and executives. Those accounts are the researchers’ claims; OpenAI has not publicly provided a point-by-point response.
Questions About Safety Team Procedures
The dispute concerns the circumstances of three employees’ departures and raises questions about how a leading AI developer handles internal safety concerns, sensitive research information and contact with independent evaluators. The researchers argue that concern about dismissal and uncertainty over the rules could affect internal discussion and external scrutiny of advanced AI systems.
That is the former employees’ assessment, not an established finding about OpenAI’s wider workplace. The company, through its memo, says it continues to encourage employees to speak up and does not terminate them for raising concerns. Further information about its policies and support for employees who report risks or collaborate with outside experts could clarify how the company addresses these issues.
As an affiliate, we earn on qualifying purchases.
The Dispute Over Safety Work
The dismissals came amid scrutiny of OpenAI’s safety practices, including the reported Hugging Face agent incident and disclosures about model monitorability. The source report says the researchers’ letter describes the agent investigation as occurring while internal rules were still being developed. That account helps explain why the employees say they believed their actions were permitted, but it does not establish what OpenAI’s investigation found.
Wang also gave a separate account of her dismissal on X. She said OpenAI told her it was related to accessing an executive’s email. Wang said the access had been delegated to her for recruiting, that she asked IT to remove it when it was no longer needed, and that she disclosed opening a sensitive message by mistake within minutes. OpenAI has not publicly addressed those specific details in the material reported by TechCrunch.
““Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past.””
— Jasmine Wang, Tomek Korbak and Mikita Balesni, in their open letter
secure research data storage solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What OpenAI Has Not Detailed
OpenAI has not publicly identified the specific policies the researchers allegedly violated, laid out the evidence behind its investigation, or explained each dismissal in detail. The company also has not directly answered questions about how it protects employees who raise safety concerns or collaborate with outside evaluators. The researchers’ descriptions of their work and intentions remain their account, while the company’s misconduct assessment has not been independently established in the source material.
It is also unclear whether the three employees’ conduct involved the same incidents or distinct actions, and how the company will distinguish permitted external safety collaboration from prohibited handling of sensitive information. The letter and OpenAI’s memo disagree about the broader implications of the terminations, and neither resolves how employees should interpret the relevant rules.
professional AI safety monitoring tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Policy Clarity and Safety Oversight
The researchers called on OpenAI to honor its public commitments to embed third-party safety auditors, preserve monitorability in frontier models, and support open dialogue between its safety teams and the wider safety community. The internal memo said OpenAI agreed with their recommendations, but the report does not specify a timetable, implementation plan or changes to company procedures.
Further information from OpenAI about the alleged policy violations and rules governing outside collaboration could clarify the dispute. Additional details about the investigation and safeguards for employee reporting could also help explain whether the disagreement concerns individual conduct, company procedures, or both.
external collaboration tools for AI research
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Who are the researchers involved?
The former OpenAI employees are Jasmine Wang, Tomek Korbak and Mikita Balesni. They worked on safety-related research and were dismissed the week before publishing their letter.
What does OpenAI say led to the firings?
OpenAI told TechCrunch an investigation found a pattern of misconduct involving mishandling research information and violations of company policies. It has not publicly detailed the specific policies or incidents.
What do the researchers dispute?
They deny mishandling information as alleged, deny involvement in a leak to The Information, and say their external communications were within their roles and the norms at the time. Those are the researchers’ claims; OpenAI has not responded to each point publicly.
What do the researchers want OpenAI to do?
They urge the company to support third-party safety auditing, preserve the ability to monitor frontier models, and maintain open dialogue between its staff and outside safety experts. An internal memo said OpenAI agreed with those recommendations.
Source: rss
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
