Frontier AI Explained

**Executive Summary:** This report surveys frontier AI initiatives in cybersecurity—Anthropic Mythos/Glasswing, OpenAI GPT-5.5-Cyber/Daybreak, Microsoft MDASH, Google CodeMender, and open-source models—describing their capabilities, access restrictions, and roles in vulnerability discovery, remediation, and security automation, and warns that operational governance, access controls, and safeguards are required because the same capabilities that help defenders can be misused by attackers.