News
Project Glasswing: Securing critical software for the AI era
First reported anthropic.com
Page published
Earliest dated coverage: 7 Oct 2026 · First observed: 7 Oct 2026 · Latest dated coverage: 7 Oct 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Anthropic loosening safety classifiers on a frontier cyber model that can autonomously find and exploit vulnerabilities raises the stakes of both defensive uptake and potential misuse if such capabilities proliferate beyond vetted organizations.
Anthropic is merging Project Glasswing into an expanded, tiered Cyber Verification Program that gives vetted security professionals access to its advanced cyber LLMs (Claude Opus 5.5, Sonnet 5.5, and Mythos 5.1) with reduced blocking classifiers and fewer guardrails. Anthropic says Mythos significantly accelerates vulnerability discovery — having already found thousands of high-severity vulnerabilities across major operating systems and browsers — and it limits broader access because of the technology's dual-use potential for abuse by threat actors.