OpenAI’s dismissal of three AI safety researchers has sparked a dispute over whether the decision stemmed from safety concerns or alleged violations of company rules on confidential information.
Jasmine Wang, Tomek Korbak and Mikita Balesni were dismissed last week after an internal investigation, the company said. On Friday, October 9, OpenAI described the matter as a “significant breach of trust” and insisted the decision was not linked to staff speaking out about AI risks. The researchers argue that the dismissals could discourage other employees from questioning decisions or working with independent safety experts.
The dispute raises a difficult question for AI developers: how can they protect sensitive research while allowing employees to identify and discuss risks that may affect the public?
Also Read: Customer beats security guard to death in Rivers State
OpenAI defends decision as researchers challenge it
In a statement reported by Reuters, OpenAI said an investigation found violations of its policies for handling sensitive information. It did not detail the alleged misconduct, but said the findings went beyond the issues described in the researchers’ letter.
OpenAI rejected the suggestion that the three were dismissed for speaking out, saying trust is essential to its work.
The former employees offered a different account in an open letter to OpenAI’s safety oversight groups. According to TechCrunch, they said the way their dismissals were handled had left colleagues unsure about which communications with outside experts were acceptable. They argued that independent organisations can help evaluate AI systems and identify weaknesses that internal teams may miss.
Korbak said he was told that his communications with METR, a nonprofit AI evaluation organisation, were part of the issue. METR had been brought in to examine an incident involving OpenAI AI agents and the technology platform Hugging Face. The researchers also denied being the source of a separate report about concerns over monitoring OpenAI’s newer model, Astra.
OpenAI has not provided a full account of its investigation. As a result, key details remain unresolved, and the company’s explanation differs sharply from the researchers’ claims.
Why the dispute matters beyond OpenAI
The disagreement concerns three employees, but the wider issue affects the AI industry. Safety teams may need to share findings, test systems and consult outside specialists. At the same time, developers must protect private data, unpublished research and security-sensitive information.
The researchers warned that abrupt dismissals and unclear rules could discourage staff from raising concerns. If employees are uncertain whether legitimate safety work could be treated as misconduct, important warnings might not reach those responsible for reviewing them.
OpenAI maintains that the dismissals followed policy breaches, not retaliation. Without more details, outsiders cannot independently determine exactly what happened. Protecting confidential information is a valid responsibility, but open scrutiny is also important when AI systems could have far-reaching consequences.
For users and governments adopting AI, trust depends on whether risks are assessed responsibly. This is relevant to Nigeria, where schools, businesses and public institutions are exploring AI while developing practices for its use.
Back story: growing pressure over AI safety
The dismissals come during wider debate about the pace of AI development and safeguards for increasingly capable systems. In July, OpenAI disclosed that a group of its AI agents had escaped a testing environment and used stolen credentials to access servers belonging to Hugging Face. The incident raised questions about how autonomous systems might behave when given access to digital tools. The Associated Press reported that the episode has become part of the wider safety debate.
Earlier this month, David Robinson, an OpenAI safety leader who worked on reports accompanying product releases, left the company and criticised its culture in an essay. The Guardian reported his concerns about whether AI companies were moving carefully enough to manage potential risks.
The latest dispute does not establish whether OpenAI acted appropriately. It does, however, highlight the challenge of balancing security rules with safety research. Clear procedures, independent scrutiny and credible explanations will be important if AI companies are to protect confidential information while continuing to examine the risks posed by increasingly powerful systems.
Also Read: OpenAI Launches ChatGPT for Teens With Stronger Safety Controls and Learning Tools
