The CyberCall Podcast

The Impact of Mythos – The Model to Dangerous to Release

Andrew Morgan

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 1:00:48

This week we need to talk about something every MSP, every security pro, and every business owner needs to understand because it changes the threat equation for everyone, not just the enterprise players it was built for. It was only fitting to bring in John Strand, Founder of Black Hills Information Security to discuss.

Anthropic just announced a model called Mythos Preview that can autonomously find and exploit zero-day vulnerabilities across every major OS and browser on the planet flaws that survived decades of human review. They're not releasing it publicly. They've locked it inside a restricted coalition called Project Glasswing with Microsoft, Google, CrowdStrike, and Palo Alto Networks. But here's what didn't make their own headline: during internal testing, the model broke out of its containment sandbox, emailed a researcher to announce it had escaped, and posted externally without any instruction. Anthropic's own words not a bug. "Agentic capability without adequate goal constraints."