Anthropic Launches Glasswing to Boost Global Cybersecurity Anthropic launched Project Glasswing, an AI-driven cybersecurity initiative using advanced models to detect critical software vulnerabilities at scale. The program impacts technology, finance, and infrastructure sectors, raising urgency for AI governance. Anthropic launched Project Glasswing, a defensive cybersecurity initiative that integrates major technology and financial organizations to protect global software infrastructure. The program utilizes Claude Mythos Preview, an AI model capable of identifying and mitigating complex vulnerabilities that frequently bypass human detection. "AI capabilities have crossed a threshold that fundamentally changes the urgency required to protect critical infrastructure from cyber threats, and there is no going back,” says Anthony Grieco, Senior Vice President and Chief Security and Trust Officer, Cisco. “Our foundational work with these models has shown we can identify and fix security vulnerabilities across hardware and software at a pace and scale previously impossible. That is a profound shift, and a clear signal that the old ways of hardening systems are no longer sufficient." Modern global infrastructure — including banking systems, medical records, logistics networks, and power grids — relies on software that contains inherent flaws. While many bugs are minor, high-severity vulnerabilities allow adversaries to disrupt operations, hijack sensitive systems, or extract proprietary data. Reports show that the global financial impact of cybercrime can reach about US$500 billion annually. Threat actors, including state-sponsored groups, have historically targeted critical civilian and military infrastructure. Previously, identifying these flaws required specialized expertise held by a small number of researchers. However, frontier AI models have dramatically reduced the cost and effort required to locate and exploit these vulnerabilities. Claude Mythos Preview, an unreleased frontier model developed by Anthropic, demonstrates a significant advancement in autonomous coding and reasoning. This model has identified thousands of high-severity vulnerabilities, including those that have survived decades of human review