OpenAI defended its decision to fire three AI safety researchers after an internal investigation found violations of policies governing sensitive information
The terminated employees — Tomek Korbak, Jasmine Wang and Mikita Balesni — alleged that their dismissals threaten the company’s culture of open debate and independent safety oversight.
In a statement posted on X, the ChatGPT parent company pushed back against the allegations, arguing that the researchers committed a "significant breach of trust" beyond what they described in a letter to the company’s safety oversight bodies.
Benzinga previously contacted Korbak, Wang, and Balesni for comment, but did not receive a response.
The company said the decisions were unrelated to the researchers’ safety concerns.
"These decisions were not about them raising safety concerns," OpenAI said. "We have not and do not terminate any of our employees for raising concerns."
Korbak, Wang, and Balesni disputed the circumstances surrounding their departures last week. iIn a letter titled "OpenAI cannot make AI safe on its own,” they said: "Terminations such as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past.”
Researchers Dispute Circumstances Of Firings
The former employees said they had acted in good faith and within OpenAI’s working norms, arguing that collaboration with independent safety organizations was essential to evaluating advanced AI systems.
"We do not believe the path to superintelligence can be navigated safely if the people closest to the risks can no longer work in high-trust, high-bandwidth ways with each other and with third parties," they wrote.
The researchers also rejected allegations that they improperly shared information with external parties. They said policies governing some of their work were being developed in real time, making clear procedures and communication with outside organizations particularly important.
Former Researchers Urge OpenAI To Preserve Oversight
The researchers outlined three recommendations: honor commitments to independent safety assessors, preserve the monitorability of advanced AI models and maintain open communication between safety researchers and outside organizations.
They said their dismissals could lead OpenAI to restrict collaborations with groups such as METR, which evaluates AI systems, and called for independent evaluators to retain ongoing access.
"OpenAI must preserve the monitorability of frontier models," they wrote, warning that researchers still do not know how to safely develop and deploy models they cannot adequately monitor.
They also urged the company to establish clear rules for external collaboration, arguing that employees should not have to guess which activities could lead to dismissal.
"If OpenAI’s employees no longer feel like they can raise safety issues internally, or work within high-bandwidth channels with external safety organizations, we are all at greater risk that something truly catastrophic will happen," they wrote.
OpenAI said it is finalizing contracts with third-party safety assessors and expects to announce details in the coming weeks. The company agreed that preserving monitorability requires an industry-wide commitment and pointed to its published research, open-source evaluation tools and safety documentation as evidence of its ongoing work.
The company said it was "deeply sad about the outcome" and praised the researchers’ contributions, but added that it generally keeps individual employment matters private and did not believe a prolonged public exchange would be productive.
Photo courtesy: Meir