OpenAI has publicly defended its decision to terminate three safety researchers: Jasmine Wang, Tomek Korbak, and Mikita Balesni. According to an official statement posted on X, the dismissals followed a thorough investigation that uncovered a "significant breach of trust."
Policy Violations and Sensitive Information
The company asserts that the researchers violated clear policies regarding the handling of sensitive information. OpenAI's internal investigation reportedly found breaches that went beyond what the researchers disclosed in their published letter. The company emphasized that these decisions were not motivated by the researchers' safety concerns, stating that critical debates are a daily, essential part of their research process. "We have not and do not terminate any of our employees for raising concerns," the statement clarified.
The Researchers' Warning
Prior to their dismissal, the three researchers had sent a letter to OpenAI's board and safety committees. They urged the company to avoid developing AI models with reasoning processes that are incomprehensible to human overseers. Throughout September, they had also used social media to call for a slower pace in developing frontier models, highlighting safety risks amid a broader wave of cybersecurity incidents linked to AI systems.
External Oversight and Future Steps
In response to the broader concerns, OpenAI announced it is finalizing contracts with third-party safety assessors, with details expected in the coming weeks. The company also expressed agreement with the need for "monitorability" of frontier models, a field in which it claims to be investing significant resources, citing recent work on the GPT-6 Astra system card and alignment research.
