Former OpenAI researchers warn of a culture of silence
Three former safety researchers dispute OpenAI’s explanation for their dismissal and call for continued external oversight.
Three former safety researchers at OpenAI say their dismissal could have a chilling effect on colleagues who report risks associated with artificial intelligence. In an open letter, they dispute that they shared confidential information outside the company’s rules.
The researchers are Jasmine Wang, Tomek Korbak and Mikita Balesni. OpenAI previously announced that it had ended its collaboration with three employees following an internal investigation into the handling of sensitive company information. According to OpenAI, rules had been breached and the trust required for the work had been damaged.
The three write that their dismissal and the way it was handled could discourage employees from discussing safety problems. They say their work centred precisely on collaboration with external safety groups, and that discussions had taken place with managers and board members. The letter is addressed to various safety bodies within OpenAI.
A central issue is the inspectability of advanced models. The researchers warn that developers do not know how to build and deploy systems safely when they cannot easily follow those systems’ reasoning processes. They therefore ask OpenAI to continue monitoring models and to enable collaboration with independent safety researchers.
According to TechCrunch, OpenAI has not publicly released a complete dossier on the dismissals. The company told TechCrunch that an internal memo acknowledges the three researchers’ contributions to AI safety, but that, according to the memo, the dismissals were not the result of expressing safety concerns. The conflict has therefore not been independently resolved.
The issue touches on a broader debate about oversight of companies developing the most powerful AI models. Internal expertise can identify risks early, but companies must also protect their trade secrets and security. The available sources do make clear that there is a conflict over transparency and safety work, but not exactly what information was shared or which internal procedures were breached.
One story, several perspectives
What is established
- Three OpenAI researchers were dismissed.
- They say their safety work and contacts with external researchers played a role.
- OpenAI says that rules concerning sensitive information were breached.
Left
Arguments Powerful AI companies should not decide for themselves how much oversight there is. Employees who report safety risks should receive protection, and independent oversight should be mandatory.
Values Public safety, transparency, employee protection and democratic oversight.
Consequences More rules and external audits could slow development, but this approach holds that they reduce the risk of commercial pressure overriding safety warnings.
Centre
Arguments AI companies need room to protect trade secrets, but must at the same time provide clear procedures for internal reporting and controlled collaboration with external experts.
Values Proportionality, careful institutions and a workable balance between innovation and oversight.
Consequences Clear reporting channels and accountable safety committees can restore trust without making all technical information public.
Right
Arguments A company must be able to hold employees to account when they handle confidential information improperly. Safety policy works only when responsibilities and ownership rights are clear.
Values Accountability, legal certainty, entrepreneurial freedom and protection of investment.
Consequences This approach holds that overly broad whistleblower protection could encourage abuse and leaks; too little protection, by contrast, could drive good staff away.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The core facts have been confirmed by TechCrunch and the public letter from the three researchers. The contents of OpenAI’s internal investigation remain partly unknown and have been presented as uncertain.
- confirmed OpenAI dismissed three safety researchers. — TechCrunch names Jasmine Wang, Tomek Korbak and Mikita Balesni and describes their dismissal. source
- confirmed The researchers published an open letter about their dismissal and AI safety. — The letter is publicly available on Mikita Balesni’s website. source
- confirmed OpenAI says the dismissals were related to the handling of sensitive information. — TechCrunch describes OpenAI’s explanation and the internal memo. source
- confirmed The precise information shared has not been made public. — The reporting consulted does not contain a complete public investigation dossier. source
Editor's note
The dismissal of three researchers, their open letter and the researchers’ and OpenAI’s conflicting explanations are certain. The precise confidential information and internal decision-making have not been made public.Sources
- Fired OpenAI safety researchers dispute misconduct claims — TechCrunch
- OpenAI cannot make AI safe on its own — Mikita Balesni