Anthropic · 2026-06-02 · major
Anthropic Expands Project Glasswing to ~150 New Critical-Infrastructure Partners Across 15+ Countries — Power, Water, Healthcare, and Communications Get Claude Mythos Plus Opus 4.8 Claude Security
Anthropic is rolling Project Glasswing out to roughly 150 new critical-infrastructure organizations across 15+ countries after the initial 50 partners found 10,000+ high- and critical-severity flaws with Claude Mythos. Claude Security on Opus 4.8 scans codebases and proposes patches.
Anthropic scales its Mythos-powered defender program from 50 to ~150 critical-infrastructure partners across 15+ countries.
What is it?
Project Glasswing is Anthropic's restricted partner program that puts Claude Mythos Preview — its cyber-defense model — into the hands of vetted security teams for vulnerability research. The June 2 update expands the program with roughly 150 new partners spanning power, water, healthcare, communications, hardware, and software vendors, and bundles a Claude Security product built on Opus 4.8 for code scanning and patch suggestions.
How does it work?
Partners run Claude Mythos Preview against their own codebases and dependency graphs; the model chains reasoning over large repos to surface exploit-chained bug patterns. The new Claude Security product wraps Opus 4.8 with workflows for triage, code review, and patch drafting. Anthropic keeps access gated to organisations whose compromise could cascade into wider harm.
Why does it matter?
Anthropic expects other labs to ship Mythos-class capabilities within 6 to 12 months, so it is racing to harden the operators whose breach could affect 100M+ users each. Defenders get earlier visibility on critical bugs; the program is also Anthropic's blueprint for how restricted-access AI products can ship before the offense side catches up.
Who is it for?
vulnerability-research and AppSec teams at critical-infrastructure operators and their suppliers
Try it
https://claude.com/product/claude-security