AI Cuts Code Vulnerability Review Time From 8 Years To 8 Weeks
In-house counsel leading cybersecurity, technology product, and enterprise risk functions for organizations with large codebases must update vulnerability management and patch cycle policies to account for AI-driven acceleration of code review timelines.
This episode features Cisco Senior Director of Technology Policy Eric Wenger sharing results from Project Glasswing and Project Daybreak, where frontier AI models reviewed 1.8 billion lines of code in 8 weeks, a task previously estimated to require 8 years, with a false positive rate of roughly 3%. Wenger outlines ongoing policy tensions around providing cybersecurity defenders early access to advanced AI models versus accelerating commercial model releases, and notes emerging government initiatives including an AI vulnerability clearinghouse. He emphasizes that while AI drastically speeds vulnerability detection, organizations must retain focus on foundational security controls including multi-factor authentication and least-privilege access, and adjust internal patch cadence policies to align with faster AI-powered detection timelines.