\cleardoublepage \part{The Blue Team's AI Team} \setbookpart{The Blue Team's AI Team}

Part III: The Blue Team's AI Team

The blue team uses AI and defends it. Analysts query, draft and summarise with assistants; agents enrich, investigate and in places act; and everything these systems read may have been written by the adversary. This part treats the team's AI as part of the estate: how analysts use it, how an agent team is governed by mandates, how its conclusions are verified when the reviewer depends on it, and how it is protected.

Chapters in this part:

• Chapter 12, The Analyst's Own Use of AI: Querying, Drafting and Summarising in Modes 1 to 3, treats the analyst's assistants and the data they may see.
• Chapter 13, An AI Agent Team for the SOC: Skills, Mandates and Containment, treats agents with mandate manifests, identities and a stop path.
• Chapter 14, Verifying What the AI Concludes: Review for Soundness When the Analyst Depends on the AI, treats review that stays real at volume.
• Chapter 15, Defending the Defenders' AI: Poisoned Telemetry, Hostile Tickets and the SOC as Tier 0, treats the SOC's AI inside its own attack surface.

Blue team functions in this part. The part concerns all five functions, with B2 Detect, B3 Investigate and B4 Respond most of all.

The thread from the series. The Transformed Security Architect described an architect's AI team and the review of what it proposes; AI Platform and Security showed that prompt injection is a property of the architecture. This part applies both to the SOC.

What you will learn in this part:

• How to govern analysts' use of assistants without driving them to shadow AI
• How an agent team is introduced, bounded and withdrawn
• How to measure whether review is real
• How to protect the SOC's AI and platform as tier-0

Part IV places the team in the organisation.