ANTHROPIC’S AI FOUND 10,000 CRITICAL BUGS IN CRITICAL INFRASTRUCTURE — NOW IT IS EXPANDING TO 150 MORE ORGANIZATIONS Anthropic’s Project Glasswing, a cybersecurity initiative that deploys the Claude Mythos AI model to scan critical infrastructure codebases for vulnerabilities, is expanding to 150 additional organizations across more than 15 countries. The expansion adds sectors underrepresented in the initial rollout, including power, water, healthcare, communications, and hardware manufacturing. The program launched in April with 50 initial partners including the U.S. government and has already produced results that are hard to ignore. Partners collectively identified more than 10,000 high or critical-level security flaws in their systems. Cloudflare alone flagged 2,000 bugs across its most critical infrastructure paths, 400 of which were rated high or critical severity, with a false-positive rate the company described as better than human testers. Major initial partners include Apple, Nvidia, Microsoft, CrowdStrike, and Palo Alto Networks. New partners joining the expanded program will need to meet security screening requirements before gaining access to the Mythos model. The scale of what Glasswing has uncovered raises uncomfortable questions about how exposed critical infrastructure already was. Power grids, water systems, and hospital networks apparently carrying thousands of undetected critical vulnerabilities is not a reassuring picture. Anthropic is positioning Claude as the solution. Given the early numbers, it is difficult to argue with the approach. Keywords: Anthropic Project Glasswing, Claude Mythos, critical infrastructure cybersecurity, AI security 2026