Anthropic expands Project Glasswing to 150 more organizations

The cyberdefense program gives approved partners access to Claude Mythos Preview to scan codebases, identify vulnerabilities, and support patching work across critical software systems.

Anthropic is expanding Project Glasswing to approximately 150 new organizations using Claude Mythos Preview for software vulnerability scanning and patching support.

Anthropic is expanding Project Glasswing to approximately 150 new organizations, extending access to Claude Mythos Preview for approved partners working to identify and address software vulnerabilities.

The expansion, announced on 2 June 2026, follows an initial cohort of around 50 partners that began using Claude Mythos Preview in April to scan codebases for vulnerabilities. Anthropic says those partners have so far found more than 10,000 high- or critical-severity security flaws.

The new group includes organizations based in more than 15 countries. Anthropic says many provide critical infrastructure, while others maintain codebases relied on by governments, companies, nonprofits, and other organizations.

The expansion adds sectors that were not strongly represented in the first cohort, including power, water, healthcare, communications, and hardware. Each organization will need to meet Anthropic’s security requirements before gaining access.

For developers, software teams, cybersecurity researchers, and education providers teaching AI and cyber skills, the program shows how frontier AI systems are moving into code review, vulnerability detection, patching, penetration testing, and secure software development.

Claude Mythos Preview used for vulnerability scanning

Project Glasswing is Anthropic’s collaborative effort to secure software systems by giving selected partners access to Claude Mythos Preview.

Anthropic says the first partners have used the model at scale, shared information and working methods with other participants, and worked with third parties to triage findings.

The company says the expanded cohort includes vendors, companies, and nonprofits that maintain codebases used by many other organizations worldwide.

Anthropic estimates that, for most Project Glasswing partners, a major attack on their codebase could affect more than 100 million people and have implications for global and national security.

The company says the program has prompted discussions with the software industry and governments about how AI is changing cybersecurity.

Patching becomes the main bottleneck

Anthropic says the challenge is shifting from finding vulnerabilities to verifying, disclosing, fixing, and deploying patched software.

Some Project Glasswing partners are now using Claude Mythos Preview to write patches and run pre-release checks designed to prevent vulnerabilities from being introduced.

Anthropic also says models such as Mythos Preview can be used for penetration testing, automated threat detection and response, and rebuilding legacy codebases in memory-safe languages.

The company says it is in discussions with third parties on how to scale the review and patching of vulnerabilities in open-source software.

Anthropic is also working on approaches for disclosing vulnerabilities to open-source maintainers, with the goal of making reports easier to triage and act on.

Claude Security released for code scanning

Alongside Project Glasswing, Anthropic has released Claude Security, a product that uses its latest public frontier models, including Claude Opus 4.8, to scan codebases and suggest patches.

Anthropic is also releasing, on request to trusted security teams, tools developed to help Project Glasswing partners find vulnerabilities more quickly.

The company says its longer-term goal is to support new initiatives, standards, and infrastructure for a period in which AI models have stronger cyber capabilities.

Anthropic says it expects many other AI companies to have models with Mythos-class capabilities within six to 12 months. The company says future general access to Mythos-level capabilities will require stronger safeguards to prevent misuse.

Project Glasswing will continue to expand, with Anthropic prioritizing additional essential infrastructure providers, maintainers of critical open-source software, and safety testers. The company also plans to scale its Cyber Verification Program, which would grant Mythos-class capabilities to more organizations for specific cyberdefense tasks.