Project Glasswing: Strengthening Critical Software Security for the AI Era

Anthropic is launching a large-scale initiative for the security of critical software worldwide by announcing Project Glasswing. Global technology leaders including AWS, Apple, Google, Microsoft, NVIDIA, Cisco are participating in this project.

The core driving force of the project is Anthropic's unreleased frontier model, Claude Mythos Preview. This model has coding capabilities that surpass skilled experts, and has already proven powerful enough to discover thousands of high-risk vulnerabilities in major operating systems and web browsers.

Based on the judgment that if such AI capabilities were misused they could pose a serious threat to the economy and national security, Project Glasswing focuses on redirecting this technology toward defensive purposes to protect cyber infrastructure.

To support the project, Anthropic will provide up to $100 million in usage credits for Mythos Preview, and will directly donate $4 million to open-source security organizations.