Companies
OpenAI Researchers Criticize Company for Prioritizing Profit Over Safety in Open Letter
Three safety researchers from OpenAI, Jasmine Wang, Tomek Korbak, and Mikita Balesni, were terminated on or around October 1, 2026, for allegedly sharing confidential materials with an external AI safety organization. The researchers contend that their actions were part of their professional responsibilities, particularly given Korbak's previous role as a liaison to external safety groups.
On October 8, the trio sent an open letter to OpenAI's board and safety committees, which has since been published online. The letter warns about the risks associated with the loss of “chain of thought” monitoring, a process that allows safety researchers to track AI models' intermediate reasoning. They argue that the industry lacks clarity on safely building and deploying models without human oversight.
The researchers proposed two key solutions: fostering industry-wide collaboration on model monitoring and involving independent safety auditors. OpenAI, however, has maintained that the firings were unrelated to the researchers' advocacy for safety, asserting that the terminations were due to breaches of trust.
This incident follows a pattern of safety-related exits from OpenAI, including the recent resignation of David Robinson, the safety transparency lead, who had voiced concerns about the company's internal culture and safety measures. The broader implications of this situation highlight the ongoing tension in the AI industry between the need for safety oversight and the protection of proprietary research.
FAQ
What led to the termination of the three OpenAI researchers?
The researchers were terminated for allegedly sharing confidential materials with an external AI safety organization, which they claim was part of their professional responsibilities.
What concerns did the researchers raise in their open letter to OpenAI's board?
The researchers warned about the risks associated with the loss of 'chain of thought' monitoring, which is crucial for tracking AI models' intermediate reasoning, and highlighted the lack of clarity in safely building and deploying models without human oversight.
What solutions did the researchers propose to address safety concerns in AI?
They proposed fostering industry-wide collaboration on model monitoring and involving independent safety auditors to enhance safety oversight in AI development.
How did OpenAI respond to the terminations of the researchers?
OpenAI stated that the firings were unrelated to the researchers' advocacy for safety and were instead due to breaches of trust.
What does this incident indicate about the culture at OpenAI regarding safety measures?
The incident reflects ongoing tensions in the AI industry between the need for safety oversight and the protection of proprietary research, as evidenced by previous safety-related exits from the company.