Anthropic’s Project Glasswing: 10,000+ Vulnerabilities Found in One Month, and the Patching Problem Has Never Been More Obvious

Anthropic's Project Glasswing used the Claude Mythos model to scan over 1,000 open-source projects and flagged more than 10,000 vulnerability candidates in one month, with 1,726 validated as real and 1,094 confirmed high- or critical-severity; the initiative has produced 97 upstream patches and 88 advisories while warning that AI-driven discovery outpaces current patching practices and poses misuse risks if such models are widely released.