AI Incident Response Playbook: Detection, Containment, and Forensics for AI-Layer Compromises
Most enterprise incident response playbooks assume the compromise sits at the network, endpoint, or application layer. AI-layer incidents (prompt injection in production, agent tool-call escalation, model-extraction attempts, credential theft via LLM-operated post-exploitation, data exfiltration through prompts) require a different detection signal, a different containment action, and a different forensic timeline. This playbook walks through the AI-layer incident classes the SOC should recognize, the detection signals each class produces, the containment actions that work at the AI request boundary, the forensic evidence the post-mortem needs, and the integration points with the rest of the security operations stack.