Project Glasswing Proved AI Can Find the Bugs. Who's Going to Fix Them?

Project Glasswing demonstrates how AI vulnerability discovery is outpacing the cybersecurity industry's ability to patch bugs before exploitation.

• •Mythos model achieved 72.4% exploit success rate and discovered vulnerabilities undetected for decades, including a 27-year-old OpenBSD bug
• •Less than 1% of Mythos-discovered vulnerabilities were patched, exposing a critical gap between discovery and remediation speeds
• •Defenders operate on calendar speed (~4 days per cycle) while attackers leverage AI for machine-speed autonomous attacks
• •Security programs need signal-driven validation, environment-specific risk assessment beyond CVSS scores, and closed-loop remediation without manual handoffs
• •Autonomous exposure validation platforms can compress response cycles from four days to three minutes using AI agent teams