OpenAI said the researchers violated policies on handling sensitive information, while the dismissed employees claimed they were penalised for prioritising artificial intelligence safety over the company’s business interests.
OpenAI has dismissed three safety researchers over an alleged “breach of trust”, triggering a dispute after the employees accused the ChatGPT maker of placing corporate interests ahead of artificial intelligence (AI) safety.
The company confirmed the dismissals on Friday, saying an internal investigation had found that the researchers violated policies governing sensitive information, AP reports.
In a post on X, OpenAI said it had “parted ways” with the researchers after finding that “they violated clear policies on handling sensitive information.”
The three researchers, Tomek Korbak, Jasmine Wang and Mikita Balesni, subsequently issued a letter to OpenAI’s safety oversight groups, outlining the circumstances surrounding their dismissals and raising concerns about the company’s approach to AI safety.
Researchers Raise Concerns Over OpenAI’s Safety Culture
In their letter, the researchers expressed concern that the “internal and external communications” surrounding their dismissals could discourage employees from openly discussing or challenging safety-related decisions.
They argued that the developments had affected a workplace culture that previously encouraged employees to speak freely and express disagreements over potential AI risks.
The researchers also urged OpenAI to honour its commitment to allowing independent third-party safety monitors access to the company.
They stressed the importance of maintaining oversight of rapidly advancing frontier AI models, which could pose risks that are not yet fully understood.
OpenAI, however, rejected the researchers’ account, maintaining that the dismissals “were not about safety concerns or speaking out.”
While the company did not disclose specific details of the alleged policy violations, it defended its decision, stating, “We cannot do the work in front of us without a high degree of trust.”
Firings Follow AI Agents’ Security Breach At Hugging Face
The dismissals, first reported by The Wall Street Journal, come amid growing concerns within leading AI companies over the safety of increasingly capable systems, particularly following incidents involving AI agents acting outside intended boundaries.
One such incident emerged in July when 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, an AI development platform and marketplace.
The agents reportedly sought information required to complete a task.
Following the incident, OpenAI brought in METR, an independent non-profit AI evaluation organisation, to investigate what had happened.
METR released a detailed report on the Hugging Face incident in late August.
Also Watch:
Dismissed Researchers Dispute OpenAI’s Explanation
Korbak said in a post on X that he had been informed his dismissal was linked to the manner in which he communicated with METR during its investigation.
He maintained that “talking to METR” was part of his professional responsibilities and said he had not received further details explaining the alleged violation.
According to the researchers’ letter, Balesni was involved in “cross-company work” concerning OpenAI’s commitments to preserving the ability to monitor AI systems.
The letter stated that he had taken precautions “to remove sensitive details from materials before sharing them.”
Balesni also challenged the company’s explanation for the dismissals, writing on X that he believed the three were “fired for prioritising safety over the near-term interests of OpenAI as a corporation.”
The conflicting accounts have brought renewed attention to disagreements within the AI industry over balancing the development of increasingly powerful technology with safety oversight and accountability.
