Project Glasswing Expands

Project Glasswing is Anthropic's collaborative project to protect critical software worldwide. Initial partners used Claude Mythos Preview to achieve the result of discovering over 10,000 high-risk security vulnerabilities.

Through this expansion, approximately 150 new organizations will join the project. These span critical infrastructure sectors directly tied to national and global security—including power, water, healthcare, telecommunications, and hardware—and are distributed across 15+ countries.

Anthropic aims to build security defense systems in preparation for a future where AI models with powerful cyber capabilities become widespread, and it performs the following two roles:

• Supporting security defenders: Providing products like Claude Security and vulnerability detection tools to help industry adapt.
• Accelerating patching and security processes: Going beyond vulnerability discovery to support the entire process of disclosing, fixing, and deploying patches.

The current bottleneck in cybersecurity lies in the speed of verifying and patching the vast number of vulnerabilities found by AI models. Anthropic aims to address this problem by using Mythos-class models to support patch writing and pre-release checks.