AI in Cyber Operations: From Scripts to Autonomous Systems
AI's role in cyber operations is not just about speed anymore. It's about scale and autonomy, reshaping offensive capabilities.
Recent developments in AI-driven cyber operations reveal a shift from traditional speed-focused strategies to those emphasizing scale and autonomy. Anthropic's Claude Mythos Preview disclosed a model capable of independently identifying and exploiting vulnerabilities, including a critical 17-year-old flaw in the FreeBSD NFS server (CVE-2026-4747). This model's capacity to execute multi-stage attacks autonomously underlines a paradigm shift in offense economics.
The Mythos Preview surfaced high-severity vulnerabilities across major systems, chaining them into novel attacks with minimal human input. Anthropic's defensive coalition, Project Glasswing, involves major players like AWS, Google, and Microsoft, focusing on countering these AI-driven threats.
US government actions mirror this shift, with the FY 2026 NDAA directing the DoD to develop AI cybersecurity frameworks, reflecting a move from experimental to operational phases. AI enhances reconnaissance and attack generation, challenging defenders still reliant on human-paced decision-making.
The emerging AI cyber doctrine prioritizes speed over stealth, adaptive systems over static controls, and probabilistic defense over zero-loss ideals. As AI compresses time, cost, and expertise constraints, defenders must adjust or risk being outpaced by machine-speed offenses.
