Project Glasswing: AI Finds What Decades of Human Review Missed

Anthropic + 12 founding members including Microsoft โ€” Apr 7, 2026
1000s
zero-day vulns found
by Claude Mythos Preview
$100M
usage credits committed
by Anthropic
40+
orgs using Mythos Preview
on critical infrastructure
๐Ÿ” What It Does

"Claude Mythos Preview" โ€” a limited-release frontier model โ€” scans source code for vulnerabilities that survived decades of human review across every major OS and browser.

Not static analysis. Not fuzzing. An AI that reasons about code the way an expert security researcher would, but at machine scale.

Key Quote (Daniela Amodei)

"AI cyber capabilities at this level will proliferate over the coming months, and not every actor who gets access to them will be focused on defense. That's the gap Glasswing is built to close."

๐Ÿค Founding Coalition
Anthropic AWS Apple Broadcom Cisco CrowdStrike Google JPMorganChase Linux Foundation Microsoft NVIDIA Palo Alto Networks
We're a founding member

Microsoft evaluated Mythos Preview against our own CTI-REALM benchmark. More on the next slides.

Sources: anthropic.com/glasswing ยท MSRC Blog ยท CTI-REALM Benchmark