Google Chrome AI Security: AI Agents Fix Vulnerabilities

### Executive Summary This article describes Google’s deployment of AI agents across Chrome’s security lifecycle to scan code, analyze historical commits and CVEs, triage findings with Critic AI, generate patch candidates (including integrations with DeepMind and Project Zero tooling), and create automated regression tests; it highlights reported outcomes (1,072 vulnerabilities discovered, a long-hidden sandbox escape, and >20 issues blocked from production) while noting the need for human oversight and recommending standard security best practices.