Microsoft Turns to Anthropic’s Mythos to Improve Cyber Defense

Microsoft is integrating Anthropic's Claude Mythos Preview model into its Security Development Lifecycle (SDL), embedding AI directly into code writing and testing stages. Access was granted through Anthropic's Project Glasswing program, which provides limited access to select tech firms for cybersecurity research. The approach shifts from static analysis to dynamic AI-driven vulnerability detection, enabling earlier identification and remediation of complex flaws. Microsoft also plans a multi-model scanning platform entering preview in 2026. However, the same AI capabilities that improve defense also compress the window between vulnerability discovery and exploitation, and human oversight remains essential for novel threat categories.