Anthropic Deploys Mythos to Harden Critical Software

Anthropic announced Project Glasswing, a defensive-security consortium that gives more than 40 partner organizations private access to an unreleased frontier model — Claude Mythos Preview — to find and help patch software vulnerabilities. Anthropic says Mythos, a general-purpose model with strong coding and reasoning skills, identified “thousands of zero-day vulnerabilities,” including many critical and decades-old bugs across major operating systems and web browsers. The model will not be broadly released; access is restricted to a coalition that includes Apple, Amazon, Google, Microsoft, Cisco, Broadcom, CrowdStrike, Palo Alto Networks, and the Linux Foundation. Anthropic is also committing up to $100 million in Claude usage credits to the effort. The initiative is explicitly defensive, framed as a way to give defenders a “head start” before these capabilities proliferate.