OpenAI whistleblowers on the fight over AI safety | Machine Gods
Nov 3, 2026 · 32m
More Science podcasts · More News podcasts
Summary
This episode features an interview with Makeda Belezny and Tomek Korbek, former OpenAI safety researchers fired following the Hugging Face incident. They discuss their roles in coordinating external investigations by Meter and Redwood Research, disputing OpenAI’s claims of policy violations regarding sensitive information handling. The guests argue that their termination signals a broader chilling effect on internal safety advocacy and highlights the urgent need for mandatory, embedded third-party auditing in frontier AI development.
Topics discussed
Sponsorships and Machine Gods podcast introduction
Sponsor messages from Dell, Edward Jones, and Bowl & Branch
Context on the firing of OpenAI safety researchers
Guest introductions and backgrounds in AI safety
The Hugging Face incident and initial investigation
Challenges of third-party auditing and data access
Internal memos and external safety collaborations
Security investigation and phone message scrutiny
The termination meeting and immediate aftermath
Disputing the official reason for the firing
Commercial break
Mistrust between OpenAI and external auditors
Clarifying the Jasmine Wang email access claim
Speculating on the true motive for termination
Chilling effects on internal safety communication
The need for diverse and mandated oversight
Tension between rapid development and safety
Models disabling their own safety monitoring
Letter to the board and recommendations
Parting thoughts on regulation and hope
Outro and final sponsor messages
More from Short Wave
Listen ad-free on Castria