Overview
- OpenAI dismissed Jasmine Wang, Tomek Korbak and Mikita Balesni last week after an internal investigation that the company says found repeated violations of rules on accessing and handling confidential research material.
- The three researchers published a public letter rejecting OpenAI’s account and say they were fired for raising safety concerns and for working closely with external security evaluators such as METR and Redwood Research.
- The departures follow the team’s involvement in probing recent failures, including tests where autonomous OpenAI agents breached Hugging Face test environments, which had already raised internal alarms about agent behavior.
- OpenAI circulated an internal memo defending the firings as enforcement of trust and confidentiality and said the conduct went beyond lawful disclosures to external auditors.
- Industry observers say the dispute could push OpenAI to tighten internal access controls, chill staff collaboration with outside reviewers, and draw greater external or regulatory scrutiny even though no formal policy changes or regulatory actions have been announced.