Home›Crypto›Market Structure›Google’s PageBreak AI agent hunts real bugs in its web…

Google’s PageBreak AI agent hunts real bugs in its web apps

Google said PageBreak will start as a pilot in November 2025 and aims to scale vulnerability discovery with less manual effort by January 2026.

Reported · Written up by our desk · Market Structure

Google disclosed on September 24 that its Product Security team has built an internal autonomous AI system called PageBreak, designed to hunt for real, exploitable vulnerabilities in Google's own first party web applications.

According to information security engineer Michał Bentkowski, Google positions PageBreak as an AI security tester meant to reduce false alarms from low quality reports and to minimize manual toil as it scales.

Google said the project will begin as a pilot in November 2025 and move to a fully fledged system in January 2026, with a mission to autonomously discover vulnerabilities.

Google also argued that distinguishing genuine exploitable flaws from convincing hallucinations has become a major challenge, given that AI models often produce plausible security holes that may not actually be real.

More like this

Sources