Google Built an AI That Hunts Its Own Security Bugs

MonitoringAI & TechSecurity incidentFirst tracked 2026-09-25Last changed 2026-09-25

Current outcome

Google disclosed PageBreak, an internal AI agent that has autonomously uncovered more than 500 confirmed exploitable security bugs in its own web applications with near-zero false positives, and plans to combine it with its CodeMender automated bug-fixing agent.

Progress timeline

1 material updates

#0109.25
Initial2026-09-25 19:16 · publication time
Google Built an AI That Hunts Its Own Security Bugs
Google disclosed PageBreak, an internal AI agent that has autonomously uncovered more than 500 confirmed exploitable security bugs in its own web applications with near-zero false positives, and plans to combine it with its CodeMender automated bug-fixing agent.
Source evidence: Decrypt

All events · Back to the feed