AI Threat Readiness Pillar 4: Detect and contain threats in real-time

Traditional security detection tools were not designed to handle AI-era threats, which move faster and create new attack surfaces through prompt injection (tricking AI by hiding instructions in its input), coding agents accessing codebases, and cloud-native AI services. The document argues that manual investigation by security analysts is too slow when the time between initial access and damage can shrink to minutes, requiring instead real-time detection with automated investigation and containment rather than human-driven responses.