Three fired AI safety researchers (Jasmine Wang, Korbak, and Balesni) jointly published an open letter formally rebutting OpenAI’s allegations that they violated sensitive information handling protocols. They asserted that communicating with external bodies like METR was within their job scope and followed established procedures. They argued the sudden dismissals undermined OpenAI’s celebrated culture of openness and triggered a severe "chilling effect," causing employees to fear retaliation and hesitate to speak out on AI risks or collaborate externally.
Cross-verified sources indicate that while OpenAI maintains the firings stemmed from policy violations rather than research content itself, the researchers emphasize that restricting communication among those closest to risk assessment hinders safe AGI development. The letter proposes three core recommendations, including allowing third-party security auditors onsite to maintain transparency, and urges the board to address this cultural crisis to ensure effective monitoring of frontier models.
This incident highlights the deep tension between commercial confidentiality and safety transparency in major AI firms. If internal candor mechanisms fail, independent oversight of frontier models may be weakened. Future developments will focus on whether OpenAI adopts the third-party audit proposal and the ripple effects on trust and regulatory compliance within the global AI safety community.