OpenAI whistleblowers on the fight over AI safety | Machine Gods

The firing of OpenAI safety researchers Mikita Balesni and Tomek Korbak highlights critical tensions between rapid AI development and the necessity for rigorous, independent oversight. These researchers, who were terminated alongside colleague Jasmine Wang for alleged policy violations, argue their actions—including coordinating with external auditors from MITRE—were essential to maintaining model safety and transparency. The situation underscores the limitations of voluntary third-party engagement, where companies often restrict access to sensitive data to mitigate legal and reputational risks. Balesni and Korbak advocate for mandated, continuous auditing frameworks similar to those in the aviation or nuclear industries to prevent the loss of "monitorability" in frontier AI systems. Ultimately, the conflict reveals a deepening cultural divide within AI labs, where commercial pressures frequently clash with the long-term safety goals required to manage increasingly autonomous and unpredictable machine intelligence.