Google deploys AI agents to secure infrastructure and automate code patching

01 What happened

Google has implemented an AI-driven security framework that analyzes code changes in real-time before they are committed. By utilizing the open-source Mantis harness, the company automates the detection, validation, and remediation of potential security flaws across its codebase.

02 Key details

• The system performs real-time pre-submit scanning on all code changes, preventing hundreds of vulnerabilities from reaching production each month.
• A triage agent verifies the reachability of vulnerable paths using abstract syntax tree parsing, achieving over 92% precision in under one minute.
• An automated bug-fix agent generates code patches based on scan results to ensure compliance with internal engineering standards.

03 Why it matters

This approach shifts security earlier into the development lifecycle, allowing for continuous, low-latency vulnerability detection and automated resolution without manual intervention.

04 Who it matters to

Cybersecurity professionals, software developers and heads of IT departments.