Definition
A structured system that coordinates multiple specialized AI agents — for example ones that explore code, find issues, validate findings, and summarize results — to automatically hunt for security flaws in software, working faster than manual code review alone. Google's Mandiant unit publicly disclosed one such production-grade system.
Why it matters
This shows AI-driven vulnerability hunting is moving from experimental research into daily production use at major security vendors, changing the speed at which both defenders and attackers can find weaknesses in software.