HeadsUpAI

Anthropic Expands Project Glasswing to Secure Global Critical Infrastructure

Anthropic is expanding Project Glasswing, its defensive cybersecurity coalition, to approximately 150 organizations across 15 countries. New partners gain access to Claude Mythos Preview, a frontier model designed to scan codebases for vulnerabilities. The cohort includes critical infrastructure providers in sectors like power, water, and healthcare.
New organizations
Approximately 150
Geographic reach
15+ countries
New sectors
Power, water, healthcare, and others
Vulnerabilities found to date
10,000+
Model class
Claude Mythos Preview

This expansion follows a pilot where partners identified over 10,000 critical vulnerabilities in essential software. Anthropic warns that Mythos-class models—AI capable of autonomous cyberattacks—will likely emerge within 6 to 12 months. Scaling access now aims to establish defensive norms and accelerate patching before these capabilities are released elsewhere.

While Claude Mythos Preview remains restricted, organizations can use Claude Security to scan codebases using public models. Anthropic is also releasing its internal vulnerability-finding tools to trusted security teams upon request. The company plans to scale its Cyber Verification Program to grant specific defensive capabilities to more organizations globally.

View the full update on anthropic.com
Anthropic
Anthropic
@AnthropicAI
X

We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries. Read more about this expansion and our future plans for Project Glasswing: https://t.co/QrtHSBdRbh

414retweets3.8klikes
View on X

Still wondering? A few quick answers below.

Project Glasswing is a collaborative initiative by Anthropic to secure critical software using AI. It provides vetted partners with access to advanced models to identify and fix high-severity vulnerabilities in codebases that support essential global infrastructure and open-source ecosystems.

Access now extends to approximately 150 new organizations—joining the roughly 50 initial partners from the program's April launch—across more than 15 countries. These partners include critical infrastructure providers in sectors like power, water, and communications, as well as vendors and nonprofits that maintain widely used open-source software codebases.

Anthropic expects other AI labs to release models with similar autonomous cyber capabilities within 6 to 12 months, potentially without safety safeguards. Expanding Project Glasswing now allows defenders to find and patch vulnerabilities before these powerful tools become widely available to attackers.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update