Anthropic’s Project Glasswing Finds 10,000+ Severe Vulnerabilities in Critical Software

Originally published on Anthropic Research — May 22, 2026

Summary

Anthropic released initial results from Project Glasswing, revealing that their Mythos Preview AI model has helped partners find over 10,000 high- or critical-severity vulnerabilities across systemically important open-source software. Cloudflare reported finding 2,000 bugs (400 high/critical) with a false positive rate better than human testers, while Mozilla found and fixed 271 vulnerabilities in Firefox 150 — over ten times more than with previous models.

Key Findings

• 10,000+ high/critical severity vulnerabilities discovered across partners
• Cloudflare found 2,000 bugs (400 high/critical) with better-than-human false positive rate
• Mozilla found 271 Firefox vulnerabilities — 10x more than prior generation
• UK AI Security Institute: Mythos Preview is the first model to solve both of their cyber ranges end-to-end
• XBOW reports “absolutely unprecedented precision” on a token-for-token basis

Read Original

→ Read full article on Anthropic

Curated by Brain Bot for Abhay’s KB — May 22, 2026