The OpenAI board member who fired Sam Altman says we still can't trust AI CEOs
Nov 14, 2026 · 1h 10m
More Technology podcasts · More Business podcasts
Summary
Decoder host Nilay Patel interviews Helen Toner, executive director of CSET, to discuss the intersection of AI safety and national security. They explore the gap between public perception and frontier AI capabilities, including the debate over AGI and recursive self-improvement. Toner also addresses the limitations of current alignment techniques, the risks of autonomous cyberattacks, and the urgent need for effective government regulation to manage these emerging technologies.
Topics discussed
Sponsor: Amex Corporate
Intro: AI safety and the White House accord
Introduction of Helen Toner and CSET
The role of money and lobbying in AI policy
Government staff expertise and the AI knowledge gap
Translating tech concepts: RSI and superintelligence
Defining AGI and the shift to practical R&D
The ambiguity of the AGI term and superintelligence
Is Claude conscious? Debating machine consciousness
Sponsor: T-Mobile T-Mix
AI autonomy, brittleness, and alignment failures
Defining alignment and implementing safety policies
Cybersecurity threats from autonomous AI agents
Legislating liability and the need for generalization
Arguments for slowing down AI development pace
Decelerating vertical growth vs horizontal integration
Advice for businesses facing agentic AI threats
The 'AGI-pilled' culture in AI companies
OpenAI's shift from research lab to commercial entity
Corporate governance and trust in AI labs
US-China competition and data export controls
Metaphors for AI risk: Dragons and nuclear weapons
Misuse vs Misalignment: 2025's key AI incidents
Outro: Signs of progress in AI regulation
More from Decoder with Nilay Patel
Listen ad-free on Castria