Google’s new agent security system detects tool misuse, loops and rogue behavior

We use AI agents to scan every code change that we deploy onto our infrastructure, preventing vulnerabilities from reaching production.