Anthropic Expands Project Glasswing Globally to Secure AI

Anthropic is expanding its AI security program, Project Glasswing, to help secure critical infrastructure worldwide. The company is adding approximately 150 new organizations to the program. These new partners are across more than 15 countries. Project Glasswing uses specialized AI tools to identify security vulnerabilities. The program launched in early April with about 50 partners. They have already detected over 10,000 high-severity security flaws in their codebases. This is done using Claude Mythos Preview. The new participants include organizations in sectors like power, water, health care, and communications. A successful cyberattack on many of these partners could affect more than 100 million people. Anthropic warns that "cheap, fast AI models with powerful cyber capabilities" are emerging and could be available to malicious actors soon. This expansion aims to give defenders a permanent advantage against these growing cyber threats.