Anthropic is scaling Project Glasswing to 150 new organizations across more than 15 countries, expanding its AI-powered security initiative beyond its original 50-partner cohort. The expansion, announced June 2, 2026, targets critical infrastructure in power, water, healthcare, and communications.

What Happened

Project Glasswing launched April 7, 2026, bringing together 12 major organizations including Amazon, Apple, Cisco, Google, Microsoft, and NVIDIA. Today's announcement, reported by TechCrunch, adds 150 organizations representing industries underrepresented in the first cohort: power grids, water utilities, hospital networks, and communications infrastructure.

The program uses Claude Mythos Preview, Anthropic's unreleased frontier model built for security research. Alongside this expansion, Anthropic is releasing Claude Security publicly and sharing development tools with trusted security teams.

Why It Matters

Glasswing partners have found more than 10,000 high- or critical-severity vulnerabilities to date. For most organizations in the new cohort, a successful cyberattack could affect more than 100 million people.

Claude Mythos Preview scored 83.1% on CyberGym's vulnerability reproduction benchmark, compared to Claude Opus 4.6's 66.6%. The model identified a 27-year-old OpenBSD vulnerability enabling remote system crashes, and a 16-year-old FFmpeg flaw that automated testing had missed across 5 million test runs. FFmpeg underpins video processing in nearly every major content creation and editing tool, making this discovery directly relevant to creative workflows.

Anthropic has committed $100 million in Mythos Preview usage credits to the program, plus $2.5 million to Alpha-Omega and OpenSSF and $1.5 million to the Apache Software Foundation for open-source security infrastructure.

Key Details

  • Scale: 150 new organizations join the original cohort of roughly 50
  • Countries: More than 15, with broader geographic reach planned
  • Sectors: Power, water, healthcare, communications, hardware, and open-source software vendors
  • Benchmark: Mythos Preview at 83.1% CyberGym vs. Claude Opus 4.6 at 66.6%
  • New public tools: Claude Security releasing publicly; development tools shared with trusted security teams
  • API access: Available via Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry at $25/$125 per million input/output tokens

What to Do Next

Security professionals can apply to the Cyber Verification Program at anthropic.com/glasswing for access to Claude Mythos Preview. Anthropic will publish public reports on program findings within 90 days.

For creators whose workflows depend on FFmpeg, open-source libraries, or browser-based tools, the immediate action is keeping software dependencies updated as AI-accelerated vulnerability discovery shortens the window between disclosure and active exploitation. For broader context on Anthropic's trajectory, see their recent S-1 filing.