Definition
The use of AI agents to automatically find security flaws in software at a speed and cost that was previously impossible. Where discovering a single vulnerability in a major open-source project might once have taken a human researcher weeks, AI agents can now find dozens in hours for under $1,000.
Why it matters
This compresses the window between when a vulnerability exists and when attackers can exploit it. Security and patching timelines built around the assumption of human-speed discovery are now dangerously out of date, as confirmed by both offensive research teams and binding federal directives.