Google details PageBreak AI tool for testing its web applications

Google has described PageBreak, an internal AI agent used by its product security team to test the company’s own web applications for vulnerabilities. The project began as a pilot in November 2025 and became a full project in January 2026, according to Google’s engineering account.

The team says it uses tools that aim to validate findings before sending high-confidence reports to product teams. It also keeps some unverified candidates for further investigation and tool development, recognising that automatic checks can miss complex cases.

Google reported that, as of 4 September, its scanner had identified two cross-site scripting vulnerabilities across hundreds of applications built on its high-assurance web frameworks. It said those were limited to internal applications or debug endpoints with hardening gaps. That is a description of its own environment, not a general security guarantee for other websites.

The account illustrates both the potential and limits of AI-assisted security testing: models can trace complicated paths, but Google says they can also make incorrect assumptions and require verification.

Source: Google Security Blog, September 2026.