Anthropic just changed the rules of the game with "Mythos"—a frontier-tier model that doesn't just find vulnerabilities; it reasons through a 23,000-word "Living Constitution" to decide if it even wants to help you.
In this milestone Episode 40, Damian, Troy, and Fern dive into Project Glasswing, the secret group of 12 companies given early access to this "digital nuke," and debate whether we are heading toward a future of AI-on-AI warfare. From 72% exploit success rates to AI toilets, we cover the technical, the tactical, and the hysterical.
Timestamps:
• 00:00 – Intro
• 02:15 – Project Glasswing: Why is Anthropic gatekeeping Mythos?
• 05:39 – The Stats: 72% success in generating working exploits
• 13:30 – Damian’s Deep Dive: Why BSD and zero-days are back in the spotlight
• 19:40 – The 23,000-Word Constitution: Can AI have ethics?
• 28:50 – Troy’s Reality Check: Liability and the "Self-Evolving" rule set
• 42:00 – The Great Debate: Should we bury this tech or embrace it?
• 50:00 – Fern’s Final Thought: AI toilets and the future of fiber
Podden och tillhörande omslagsbild på den här sidan tillhör
Cyber Podcast. Innehållet i podden är skapat av Cyber Podcast och inte av,
eller tillsammans med, Poddtoppen.
