Anthropic's unreleased Claude Mythos model, backed by a 12-company coalition, has already found thousands of high-severity vulnerabilities across every major OS and browser.
Anthropic announced Project Glasswing, a coalition including AWS, Apple, Google, Microsoft, NVIDIA, Cisco, CrowdStrike, JPMorganChase, and others, built around a new unreleased frontier model called Claude Mythos Preview. The model has demonstrated superhuman capability at finding and exploiting software vulnerabilities, already identifying thousands of high-severity flaws across every major OS and browser. Launch partners will use Mythos Preview for defensive security work, and Anthropic will share findings with the broader industry. Over 40 additional organizations building or maintaining critical software infrastructure have also been granted access.
Claude Mythos Preview has crossed a threshold: it outperforms all but the most elite human security researchers at finding and exploiting vulnerabilities, and it's already found thousands of high-severity flaws across every major OS and browser. This isn't a research demo — it's being deployed by 12 major tech companies right now. If your codebase or open-source dependencies haven't been scanned by a model-class tool, you're operating blind against an adversarial landscape that is about to get significantly more dangerous as these capabilities proliferate.
Run a threat-model audit on your most exposed service this week using CrowdStrike Falcon or any available AI-assisted SAST tool — prioritize anything touching OS-level syscalls, browser APIs, or memory management, since Mythos specifically flagged those surfaces.
Go to claude.ai and open a new conversation
Tags
Sources
Related
Also today
Signals by role
Also today
Tools mentioned