Fired OpenAI Researchers Warn of a “Chilling Effect” in an Open Letter

October 9 (IT Home) — Three former OpenAI researchers who were suddenly fired last week — Mikita Balesni, Tomek Korbak and Jasmine Wang — published an open letter today (the 9th). They said they worry the move will deter other employees from speaking out, producing a “chilling effect” on OpenAI’s corporate culture.

In the letter, titled “OpenAI cannot ensure AI safety alone,” the three argue that OpenAI “is not an ordinary company,” and that employees being able to work with outside organisations on safety issues without fear is an important part of OpenAI’s work culture.

The letter was posted on social media by Balesni, who previously worked on AI alignment research, and was addressed to OpenAI’s safety committees. The other two signatories, Korbak and Jasmine Wang, also previously worked on AI safety research and were likewise fired.

According to Business Insider, an OpenAI spokesperson responded that the company conducted an internal investigation and concluded that the three had violated rules on accessing and handling sensitive information.

The three researchers each said on social media that the company only told them the reasons for their dismissal verbally, and that their earlier actions all fell within the scope of their job duties.

Jasmine Wang was said to have accessed an executive’s email. She responded: “The reason the company gave for firing me simply does not hold up.”

Balesni posted that the company told him it no longer trusted him because he “communicated too much with third-party safety organisations.” “I never leaked the company’s intellectual property. My work was always coordinated with my direct manager, research leadership and the board,” he said.

Korbak also posted that the company told him verbally that he was fired because of “the way I communicated with METR.” “Let me be clear: communicating with METR was my job,” he said.

IT Home notes: METR is a non-profit research organisation that evaluates AI models and their safety risks.

In the open letter, the three worry that other employees may in future be afraid to speak their minds and do their work the way they used to — work that until last week was an indispensable part of their jobs at OpenAI. “Firing employees as suddenly as this, and notifying them so abruptly, has hit the open culture OpenAI used to be proud of with a chilling effect.”

The three also made three suggestions to OpenAI: allow third-party safety auditors into the company; ensure that frontier AI models can still be effectively monitored; and support safety researchers in maintaining “open and transparent communication” with other organisations and people in the AI safety field.

In an internal memo to employees, a head of OpenAI’s research division expressed strong support for the three suggestions. “We are very grateful for their contributions to AI safety work, and for their willingness to speak up and question existing views. We value their opinions, support their work and have given them a great deal of trust. I must be clear: the decision to dismiss them had nothing to do with them raising safety issues or speaking publicly. We have always encouraged employees to express their views and will continue to do so. We do not fire employees for raising concerns.”

Korbak also wrote on social media that he worries OpenAI will use the dismissals as a “pretext” to reduce cooperation with METR.