Project Glasswing

article Anthropic Apr 8, 2026

Anthropic announces Project Glasswing, a coalition of 12 organizations — AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks — to use Claude Mythos Preview for defensive cybersecurity. Mythos Preview is an unreleased frontier model that found thousands of zero-day vulnerabilities in every major OS and browser, including a 27-year-old OpenBSD remote crash, a 16-year-old FFmpeg bug that survived 5 million automated tests, and a Linux kernel privilege escalation chain — all discovered autonomously. Benchmark results: 93.9% SWE-bench Verified, 77.8% SWE-bench Pro, 82.0% Terminal-Bench 2.0, 83.1% CyberGym. Anthropic commits $100M in usage credits and $4M to open-source security organizations. The model will not be made generally available; safeguards are being developed for a future Opus model. Partners will share findings within 90 days, including recommendations on vulnerability disclosure, patching automation, and secure-by-design practices. Named after the glasswing butterfly (Greta oto), whose transparent wings evoke both hidden-in-plain-sight vulnerabilities and evasion of harm through transparency.

Backlinks

Anthropic's Project Glasswing: a coalition of 12 companies using Claude Mythos Preview — an unreleased frontier model — for defensive cybersecurity. The model found zero-days in every major OS and browser, autonomously. →

AI Cyber Proliferation The risk that AI-powered vulnerability discovery and exploitation capabilities spread beyond actors committed to using them safely, before defensive measures are in place. →

AI Vulnerability Discovery The capability of frontier AI models to autonomously find and exploit software vulnerabilities at a level that surpasses all but the most skilled human security experts. →

Claude Mythos Preview Claude Mythos Preview is an unreleased frontier model trained by Anthropic, announced as part of Project Glasswing. Named from Ancient Greek for "utterance" or "narrative." Not planned for general availability. →

Defensive AI Advantage The thesis that the same AI capabilities which make cyberattacks more potent also give defenders a durable advantage — if they act first. →

Cybersecurity

Project Glasswing — Anthropic (2026-04-08)

→

Project Glasswing Anthropic's announcement of Project Glasswing, a cross-industry initiative to apply frontier AI capabilities to defensive cybersecurity. →