Attackers are leveraging AI at machine speed—the first known AI-built zero-day was recently documented, and handoff time between attack stages dropped from eight hours to 22 seconds.

Francis deSouza of Google Cloud argues that defenders can counter this with a distinct advantage: deep enterprise context synthesized by AI.

The article introduces Google AI Threat Defense, a unified platform combining Gemini reasoning, Wiz cloud context, CodeMender code remediation, and Mandiant threat intelligence.

It structures vulnerability management as a continuous four-stage framework: Prepare (map assets and simulate attack paths with Wiz Red Agent), Scan & Prioritize (multi-model scanning with lightweight models for breadth and Gemini for deep risk validation), Remediate (auto-generate verified fixes via CodeMender in developer IDEs), and Monitor (deploy AI agents for runtime hunting and integrate with Google Security Operations).

A case study from Morgan Stanley reports collapsing mean time to detect threats by 99.9%, shifting from a reactive 45-minute window to proactive mitigation in 90 seconds or less.

The article emphasizes that autonomous AI agents must remain under human supervision—illustrated by Wiz Red, Blue, and Green agents for pen-testing, investigation, and remediation.

It also warns of internal risks such as shadow AI and unauthorized agents, recommending Zero Trust for AI and directing teams toward approved architectures.

The closing call: secure AI infrastructure from the ground up, not bolted on, and fight AI with AI.