The new AI model that could steal your life savings
Summary
This episode examines whether bank accounts remain safe amid the rise of powerful AI models like Anthropic’s Claude Mythos, which can identify complex software vulnerabilities. Hosts discuss how Anthropic restricted access to this model to allow companies to patch security flaws before public release, drawing parallels to the Y2K bug. Experts debate whether this proactive approach mitigates risks or if AI-driven coding introduces new, unpredictable dangers to global cybersecurity infrastructure.
Topics discussed
Intro: AI security concerns and bank CEO meeting
NPR promotions: Tiny Desk, Up First, Fresh Air
Claude Mythos: Chaining vulnerabilities and Project Glasswing
Bank safety and the Y2K comparison
Anthropic's response: Claude Security and patching bugs
Expert skepticism: Vibe coding and existing AI models
Conclusion: Industry cooperation and final thoughts
Listen ad-free on Castria