Who decides when AI is too dangerous?
Summary
Nilay Patel and Hayden Field dissect the chaotic shutdown of Anthropic’s Claude Fable 5 model following abrupt U.S. export controls. They analyze how a reported jailbreak led to a 90-minute government ultimatum, forcing Anthropic to take the model offline. The discussion highlights the erratic nature of Trump-era AI regulation, the strained relationship between Anthropic and the White House, and the broader implications for global AI competition and safety frameworks.
Topics discussed
Intro and Sponsor Ads
Introduction: The Anthropic Fable 5 Ban
Interview Start: Timeline of the Crisis
Explaining Fable 5 and Mythos Models
The Alleged Jailbreak and 90-Minute Ultimatum
Industry Reaction and Sponsor Break
Critique of Government Process and Andy Jassy's Role
Anthropic vs. Trump Administration Dynamics
Rumors, Executive Orders, and Political Posturing
Sponsor Break and Uber Interview Segment
Regulatory Frameworks and Public Sentiment
Anthropic's Safety Stance and Military AI Concerns
Conclusion and Future Outlook
Outro and Sponsor Ads
Listen ad-free on Castria