OpenAI is defending its decision to fire three safety researchers after the former employees publicly challenged the reasons for their dismissals.
Mikita Balesni, Tomek Korbak, and Jasmine Wang claim their terminations were connected to their work on AI safety and collaboration with outside researchers. OpenAI disputes those allegations, stating that an internal investigation found violations of policies governing sensitive information and a significant breach of trust.
This disagreement centers on how the company manages confidential research and collaboration with external safety organizations. The former employees maintain that their activities were consistent with accepted working practices, while OpenAI says its policies were violated.
The company has reaffirmed its support for internal safety discussions and independent evaluations. The full findings of its investigation have not been made public.
Why It Matters: The dispute brings attention to how AI developers establish rules for handling sensitive information while conducting safety research that may involve outside organizations. It also raises questions about how companies communicate those rules and manage disagreements over their application.
- The Dismissals Followed an Internal Investigation: OpenAI says the researchers violated information-handling policies and committed a significant breach of trust. The former employees dispute the findings, maintaining that their activities followed accepted company practices. OpenAI says its investigation uncovered conduct extending further than the incidents described in their public letter. The complete findings have not been released.
- AI Monitorability Faces Technical Challenges: The researchers studied how AI-generated reasoning traces can help identify unsafe behavior. They warn that changes in model architecture could make these traces less useful, limiting visibility into how advanced systems operate. OpenAI acknowledges the importance of preserving monitorability and supports industry coordination on the issue.
- Independent Evaluations Remain a Priority: The former employees have called for continued cooperation with external safety organizations, including meaningful access for independent assessors. They express concern that their dismissals could affect existing partnerships. OpenAI says it is finalizing agreements with third-party evaluators and remains committed to external safety reviews.
- Information-Sharing Procedures Are Contested: The researchers describe situations involving external investigations and internal email access, arguing that some procedures were still being developed. OpenAI maintains that established policies were violated. The competing accounts leave unresolved how the relevant rules were communicated and applied.
- The Effect on OpenAI’s Research Culture Is Uncertain: The former employees believe their dismissals could discourage colleagues from raising concerns or working with outside experts. OpenAI rejects that interpretation and says critical safety discussions remain encouraged. No public evidence currently establishes whether the terminations have changed employee behavior or research practices.
Go Deeper -> Fired OpenAI researchers say they were let go for ‘prioritising safety’ – BBC News
OpenAI fires 3 safety researchers in dispute over AI risks – ABC News
OpenAI defends firing AI safety researchers over alleged “breach of trust” – CBS News
