Google disclosed on September 24 that its product security team has developed an AI agent named PageBreak, designed to autonomously discover and verify exploitable vulnerabilities in Google's own web applications. The system has identified over 500 cross-site scripting (XSS) vulnerabilities across Google's first-party web applications. PageBreak began its pilot in November 2025 and fully launched in January 2026, aiming to improve vulnerability discovery efficiency and reduce manual workload. Google stated that this initiative helps address the current challenge of AI generating a large volume of low-quality security reports, and plans to connect PageBreak with CodeMender, an automated patch-writing agent, to automate vulnerability discovery and remediation.