OpenAI denies retaliation after dismissing three AI safety researchers
Artificial intelligence leader OpenAI has publicly denied claims that it terminated three prominent AI safety researchers in retaliation for raising internal concerns regarding model risks. The company asserted that the dismissals of Jasmine Wang, Tomek Korbak, and Mikita Balesni resulted strictly from a internal investigation that uncovered policy violations and a "significant breach of trust".
The controversy has renewed scrutiny over corporate governance, transparency, and internal safety culture across frontier AI laboratories.
Dispute over Misconduct vs. Internal Whistleblowing
The dismissals followed a series of public statements and an open letter to OpenAI's board in which the three researchers, alongside external academics, urged leading labs to slow down frontier model deployments that diminish system monitorability.
While former employees and industry observers characterized the terminations as an effort to stifle internal dissent, OpenAI issued a public statement clarifying its position. The company stated that its internal review uncovered a pattern of unauthorized communications and data handling that extended beyond routine engagements with third-party evaluation groups.
OpenAI explicitly stated that the personnel decisions "were not about raising safety concerns or speaking out," emphasizing that internal debate over research risks is actively encouraged.
In a joint response, Wang, Korbak, and Balesni denied mishandling sensitive information outside established protocols, warning that the abrupt firings have created a "chilling effect" among remaining staff working on alignment.
Third-Party Safety Audits and Ongoing Departures
The dispute centers in part on how internal researchers interact with external safety benchmark organizations. Korbak indicated that his dismissal involved communications with Model Evaluation and Threat Research (METR), an independent non-profit auditing AI risk models.
The firings coincide with broader instability across OpenAI's safety division. David Robinson, who led transparency initiatives within the safety team, resigned shortly after the terminations, publicly citing concerns over organizational culture and communication barriers.
OpenAI reiterated that collaboration with independent safety auditing groups remains integral to its deployment strategy and noted that formal contracts with external evaluation partners are being finalized.
Escalating Debate Over Frontier AI Governance
The conflict underscores the challenge facing AI labs as autonomous agents and advanced reasoning architectures gain capabilities. As regulatory bodies in the U.S. and Europe tighten reporting mandates on frontier AI models, the balance between protecting proprietary enterprise IP and maintaining open channels for safety research remains a focal point for the industry.
