Part VII · The program

One assessment is a snapshot. A program is what keeps the answer true next quarter. Build it, run it, and eventually retire it, with evidence at every stage.

Each part of this book splits the same three ways: how it works, how it breaks, and how you hold it. The attack and the control for one layer sit together, because that is how you actually meet them.

Build it

VII.1 The secure AI SDLC The software development lifecycle for AI systems, stage by stage - what to look for, how to do it, what to implement, and the threats at each - anchored in NIST SSDF SP 800-218A. runbook · 27 min read VII.2 The Agent Development Lifecycle (Agent DLC) Agents need their own lifecycle - evaluation-first, not code-first. Nine build stages, what to secure at each, and how to harden agentic systems with MCP. runbook · 14 min read

Run it

VII.3 Detection, IR & forensics for AI What to capture, what to detect, how AI incident response differs, and discovering shadow AI across the organization. defense · 13 min read VII.4 Finding the AI you do not know about The five independent discovery feeds for shadow AI - egress, cloud control plane, developer surface, identity, and money - and how to turn findings into an inventory with owners. defense · 5 min read

Retire it

VII.5 Retirement & decommissioning Retiring an AI system: revoking identities, crypto-shredding weights, proving a vector store forgot, and keeping exactly the evidence you are still obliged to keep. defense · 8 min read