Ex-OpenAI staff say they were fired for raising safety concerns

Safety researchers Mikita Balesni, Tomek Korbak and Jasmine Wang levelled the accusations on Thursday after their dismissals last week. The trio warned of a “chilling” effect on other staff concerned about the frontier technology’s risks.

OpenAI said they were let go due to misconduct involving the mishandling of company information.

“I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation,” Balesni said in a post on X.

Korbak expressed a similar sentiment, expressing his belief that he had been terminated for raising concerns that OpenAI was “losing the ability to monitor what AI agents think, one of our best tools for catching when they misbehave”.

In an open letter published in tandem with their social media posts, Balesni, Korbak and Wang expressed concerns that their dismissal would make their former colleagues “afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI”.

“We could raise safety concerns and disagree openly, and were encouraged to draw on the expertise of independent safety organizations,” they said in the letter. “This is part of what made OpenAI special, and why we are immensely proud to have been part of the team.”

The three added that “if the people closest to the risks can no longer work in high-trust, high-bandwidth ways with each other and with third parties”, then AI could not be developed safely.

“It is that culture we are writing to defend,” they stated.

The trio denied violating company policy, saying any engagement with external safety experts had been in line with the mandate of their roles.

“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 said.

“Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past.”

In a statement, OpenAI said it had uncovered “a significant breach of trust” by the former researchers that went beyond what was described in their letter and said it stood behind the decision to fire them.

“We want to be very clear that these decisions were not about raising safety concerns or speaking out,” the San Francisco-based company said.

“Safety and research debates happen every day at OpenAI, often spirited and highly critical. We actively encourage these discussions and consider them essential to making the right decisions. We cannot do the work in front of us without a high degree of trust. We will continue to be extremely forgiving of our team making good-faith mistakes.”

OpenAI added that it was saddened by the outcome.

“We appreciated Jasmine, Mikita, and Tomek’s contributions to AI safety at OpenAI and their willingness to speak up and challenge ideas,” the company said.

“We championed their voices, supported their work, and placed enormous trust in them. These decisions were not about them raising safety concerns. We have always encouraged that and always will.”

The ex-employees’ intervention comes as OpenAI and other AI companies are at the centre of intense debate about how to ensure the rapidly evolving technology does not expose humanity to catastrophic harm.

The risk of cutting-edge AI models escaping human control has been in the spotlight since July when it emerged that autonomous agents generated by OpenAI had hacked the software start-up Hugging Face.

OpenAI and its chief rival, Anthropic, have urged the international community to work together on a coordinated slowdown in AI development.

However, that suggestion has been rebuffed by the two leading AI powers, the United States and China.

Last month, OpenAI, Anthropic, Google, Meta, SpaceXAI and Nvidia endorsed a voluntary accord calling for stronger internal safeguards and the engagement of external auditors to better manage safety risks.

The accord, announced by US President Donald Trump, drew a mixed response from AI safety advocates, many of whom criticised its nonbinding nature.

OpenAI has announced a number of steps to mitigate risks in the absence of formal regulation, including scrapping the planned release of its latest-generation model, GPT-6.1 Astra.

The ChatGPT creator said it decided not to go ahead with the release after its model failed to meet company standards for acting in accordance with human wishes.