Anthropic’s Mythos finds 10,000 critical software flaws

Anthropic says Project Glasswing used Claude Mythos Preview with about 50 partners to uncover more than 10,000 high- or critical-severity vulnerabilities.

The findings included major issues at Cloudflare, Mozilla, and other organizations, showing AI can dramatically scale bug discovery.

Separate scans of 1,000+ open-source projects also found thousands of vulnerabilities, but only a small portion have been patched.

The results suggest human patching capacity and safety controls are now the main bottlenecks before more cyber-capable models are released.