Summary
Anthropic’s first public Project Glasswing update says Claude Mythos Preview and roughly 50 partners have already found more than 10,000 high- or critical-severity vulnerabilities across critical software. The company argues the bottleneck has shifted from finding bugs to verifying, disclosing, and patching them fast enough.
What changed
Anthropic published its first Project Glasswing progress update with aggregate vulnerability findings, open-source scan results, and a new cyberdefense tooling package for qualifying customers.
Why it matters
This is a strong signal that frontier AI is already reshaping defensive security workflows, not just red-team benchmarks. The practical constraint is now disclosure and patch capacity, which raises the strategic value of triage, patch automation, and secure deployment hygiene.
Evidence excerpt
Anthropic says Claude Mythos Preview and its partners have found more than 10,000 high- or critical-severity vulnerabilities, and that software security is now limited more by verification, disclosure, and patching than by discovery.