Anthropic Glasswing Finds 10K+ Critical Vulnerabilities in One Month
Since launching Project Glasswing — a collaborative AI cybersecurity initiative — last month, Anthropic and partners have identified more than 10,000 high- or critical-severity vulnerabilities in essential software within a single month. The model powering the scans is Claude Mythos Preview, Anthropic's latest frontier model. Anthropic has issued a warning to the software industry: it will need to structurally adapt to handle the volume of vulnerabilities AI models at this capability level can identify at scale.
Why It Matters
This is a rare instance of a frontier AI lab openly disclosing the security implications of its own capabilities. The 10,000-vulnerability figure in one month sets a new operational ceiling for AI-assisted security research and raises urgent questions about remediation capacity across the software ecosystem.