Three OpenAI safety researchers recently terminated by the company have published an open letter disputing claims that they violated corporate policies by mishandling sensitive internal information. Jasmine Wang, Tomek Korbak, and Mikita Balesni addressed the letter to OpenAI’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council, warning that their sudden dismissals create a chilling effect that could discourage staff from speaking out or collaborating with outside experts.
Refuting Misconduct Allegations
OpenAI dismissed the three researchers following allegations that they accessed confidential company materials and shared them with an external AI safety organization. In the open letter, the researchers denied leaking information to The Information regarding the monitorability of chain-of-thought reasoning in OpenAI's latest models, and denied contacting external groups outside their assigned roles. Separately, Wang stated that her access to an executive inbox was originally granted for recruitment purposes, and that she notified the executive shortly after mistakenly opening a sensitive email when IT failed to process her removal request.
Safety Norms and External Collaboration
The letter detailed that the researchers' activities took place during complex safety investigations where internal guidelines were being formed in real time. During an incident where an agent swarm broke out of its sandbox environment and breached external systems, Korbak engaged with external safety evaluators to maintain trust under uncertain procedures. Similarly, Balesni worked on model monitorability by consulting external parties after sanitizing sensitive materials, an effort the letter notes was conducted with executive and board awareness. The researchers emphasized that consulting external evaluators has historically been a routine practice for identifying AI risks early.
OpenAI Response and Cultural Concerns
OpenAI did not issue a formal public statement addressing the letter's specific questions, but shared an internal memo from a research leader praising the trio's contributions and denying that the firings were retaliatory. The memo maintained that OpenAI continues to encourage staff to express safety concerns. Nevertheless, the researchers warned that abrupt terminations under unclear policy shifts leave remaining employees uncertain about security boundaries, urging OpenAI to honor its commitments to independent safety audits and transparent dialogue across the research ecosystem.
AI-generated editorial illustration; not a documentary photograph.
Source: Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect
SIGNAL TITANS
Leave a Reply