Claude Mythos: Get your AppSec game on

This article describes the leak of Anthropic marketing materials and Claude Code source material and warns that Anthropic’s Claude Mythos model greatly accelerates discovery and exploitation of zero-day vulnerabilities — potentially creating a 'vulnpocalypse' where automated agents can find, chain, and exploit bugs faster than defenders can patch them. It outlines Anthropic’s Project Glasswing (a controlled preview for select partners), reports demos and claims of thousands of findings including long-undetected bugs, and advises AppSec teams to prioritize inventory audits, patching, visibility, and layered controls to mitigate agentic AI risks.