Project Glasswing: An Initial Update
Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.
Anthropic's Glasswing interpretability project is cutting-edge AI research directly relevant.
Anthropic's Project Glasswing, with 50+ partners including Cloudflare and Mozilla, used Claude Mythos Preview to discover over 10,000 high- or critical-severity vulnerabilities in critical open-source software within one month. Partners report bug-finding rates increased by more than 10x, with Cloudflare finding 2,000 bugs (400 critical) at a false positive rate better than human testers, and Mozilla fixing 271 vulnerabilities in Firefox 150—ten times more than with the prior model. The UK AI Security Institute and benchmarks like ExploitBench confirm Mythos Preview as the strongest performer in autonomous exploit development, shifting the bottleneck from discovery to verification and patching.
Integrate AI-powered security scanning into your DevSecOps pipeline now to stay ahead of both vulnerabilities and adversarial AI capabilities.
For a Solutions Architect focused on SDLC and platform engineering, this demonstrates that AI-driven vulnerability discovery is now outpacing human capacity, forcing a rethinking of security pipelines, patch management, and CI/CD integration for critical infrastructure.
Announcements Project Glasswing: An initial update May 22, 2026 Last month, we launched Project Glasswing , our collaborative effort to secure the world’s most critical software before increasingly capable AI models can be turned against it. Since then, we and our approximately 50 partners have used Claude Mythos Preview to find more than ten thousand high- or critical-severity vulnerabilities across the most systemically important software in the world. Progress on software security used to be limited by how quickly we could find new vulnerabilities. Now it’s limited by how quickly we can…