AI Purple - Book Structure

This document describes the logical structure of AI Purple, Book 9 of The Transformed Architecture. The book is organised into seven parts with 29 numbered chapters, moving from the loop in question through its five stages, the loop's meeting with AI systems, the loop's own AI, the loop in the organisation and the people who run it, to experience and conclusions, followed by the appendices.

Table of contents

Part I: The Loop in Question (Chapters 1–4)

Before the transformation, the pressure, and the purple loop defined (defines P1 Target to P5 Verify)

Part II: The Stages of the Loop (Chapters 5–11)

The stages: target, rules of engagement, emulate, emulate continuously, observe, improve and verify

Part III: The Loop Meets AI Systems (Chapters 12–13)

The loop meets AI systems: purple for AI systems and agents, and testing the defenders' AI

Part IV: The Loop's Own AI (Chapters 14–16)

The loop's own AI: an agent team for the loop, independence and defending the loop's AI

Part V: The Loop in the Organisation (Chapters 17–22)

The loop in the organisation: coverage, measurement, operating models, decision rights, regulation and boundaries

Part VI: People and Career (Chapters 23–25)

People and career: red and blue as people, competences and the red craft, and the future of the loop

Part VII: Experience and Conclusions (Chapters 26–29)

Cases, a year in the loop, conclusions for two organisations and the verdict

Appendices and Reference

Glossary, Appendix A (code examples: scenario, rules-of-engagement, mandate, coverage, regression and evidence artefacts), Appendix B (templates and tools), references and about the author

Chapter structure

Part I: The Loop in Question

The part is introduced by part_i_the_loop_in_question.md.

Chapter	File	Title	Description
1	01_introduction.md	Introduction: The Purple Loop with AI	Introduce the book on its own terms: the loop between red and blue with AI on both sides, the thesis and what would refute it, the five stages and the independence principle
2	02_what_the_loop_did_before.md	What the Loop Did Before: Penetration Tests, Red Teams, Tabletops and Purple Sessions	Describe the loop before the transformation: penetration test, red team engagement, threat-led test, tabletop, purple session and breach-and-attack simulation, the roles, and where findings stalled
3	03_the_pressure_on_the_loop.md	The Pressure on the Loop: Machine-Speed Adversaries, Detection Decay and the Annual Test	Explain why the traditional rhythm no longer fits: shrinking exploitation windows, estates that change daily, detection decay, scarce red-team hours and finding backlogs, and what AI makes cheaper
4	04_the_purple_loop_defined.md	The Purple Loop Defined: Five Stages, Evidence and the Independence Principle	Define the frame: the five stages P1 to P5, their records, the closure rule, degrees of delegation per stage, the independence principle and the roles of red, blue, white and the purple coordinator

Part II: The Stages of the Loop

The part is introduced by part_ii_the_stages_of_the_loop.md.

Chapter	File	Title	Description
5	05_target.md	Target: Threat-Informed Scenarios, Risk and Priority Techniques	P1 Target: how scenarios are chosen from intelligence, the threat register and risk, priority techniques, the scenario library as code and AI-proposed scenarios
6	06_rules_of_engagement_and_safe_emulation.md	Rules of Engagement and Safe Emulation: Scope, Blast Radius and the Stop Path	P1 Target and P2 Emulate: the conditions under which emulation may run, authority, scope as code, blast radius, third-party consent and the stop path outside the emulating system
7	07_emulate.md	Emulate: Adversary Emulation Plans, Human-Led and AI-Assisted	P2 Emulate: the campaign, planned from intelligence and executed by people with AI assistance, and what a human contributes that AI does not
8	08_emulate_continuously.md	Emulate Continuously: Breach-and-Attack Simulation, Atomic Tests and Automated Red Teaming	P2 Emulate: emulation as a pipeline, breach-and-attack simulation, atomic tests and continuous automated red teaming; a constructive debate on whether it replaces the human red team
9	09_observe.md	Observe: Telemetry, Detection and Response Compared with Expectation	P3 Observe: comparing telemetry, detection and response with expectation layer by layer, an outcome taxonomy, ground truth from tagged events and the purple session
10	10_improve.md	Improve: From Finding to Change in Detections, Controls, Playbooks and Mandates	P4 Improve: turning a finding into the right change in a detection, control, playbook or mandate, with ownership, AI-drafted changes and review
11	11_verify.md	Verify: Re-test, Regression and Evidence	P5 Verify: the re-test, the regression case with variants, signed evidence, evidence expiry and the ratchet that keeps a closed finding closed

Part III: The Loop Meets AI Systems

The part is introduced by part_iii_the_loop_meets_ai_systems.md.

Chapter	File	Title	Description
12	12_purple_for_ai_systems_and_agents.md	Purple for AI Systems and Agents: Attacks on Models and Detections That Catch Them	The loop applied to the organisation's own AI: telemetry from the AI gateway and agent runtime, detections for agents and the path from an AI red-teaming finding to a detection and a regression case
13	13_testing_the_defenders_ai.md	Testing the Defenders' AI: SOC Agents, Playbooks and Providers' AI under Emulation	Using the loop to test the blue team's own AI: investigation and triage agents, playbooks, stop paths and a provider's AI under emulation, including injected tests in managed-service contracts

Part IV: The Loop's Own AI

The part is introduced by part_iv_the_loops_own_ai.md.

Chapter	File	Title	Description
14	14_an_ai_agent_team_for_the_loop.md	An AI Agent Team for the Loop: Skills, Mandates and Containment	Which loop functions may be given to agents and at what level: mandates, delegation profiles, identity and containment for an agent that holds offensive capability
15	15_independence.md	Independence: Verifying the Loop When AI Is on Both Sides	The central design rule: correlated failure when tester and tested share a model family or data, kinds of independence, held-out scenarios and seeded gaps; a constructive debate on whether the loop can be closed by machines alone
16	16_defending_the_loops_ai.md	Defending the Loop's AI: Scenario Libraries, Findings and Emulation Agents as Tier 0	The loop's own assets as an adversary's best map: the findings register, the scenario library and emulation agents as tier-0 assets

Part V: The Loop in the Organisation

The part is introduced by part_v_the_loop_in_the_organisation.md.

Chapter	File	Title	Description
17	17_coverage_as_a_claim.md	Coverage as a Claim: Technique Maps, Evidence Age and Detection Decay	What 'covered' means: the coverage record, evidence age and expiry, detection decay and honest 'not tested'
18	18_measuring_the_loop.md	Measuring the Loop: Cycle Time, Yield, Re-test Pass Rate and the Metrics That Mislead	Measures of the loop, not of people: cycle time, finding half-life, re-test pass rate and yield, and the metrics that mislead
19	19_operating_models_and_team_design.md	Operating Models and Team Design: In-House, Contracted, Hybrid and Sovereign Loops	Who runs the loop: in-house, contracted, hybrid and sovereign operating models, roles, role charters and procurement
20	20_decision_rights_and_the_white_team.md	Decision Rights, Authorisation and the White Team	Who may authorise, change, stop and accept: decision rights, the white team, the trusted agent and a real incident during an exercise
21	21_regulation_as_constraint.md	Regulation as Constraint: DORA, TIBER-EU, NIS 2, the AI Act, the GDPR and Employee Rights	The legal frame for emulation: DORA and threat-led testing, TIBER-EU, NIS 2, the AI Act, the CRA, the GDPR and employee rights, attributed to instrument and article
22	22_across_boundaries.md	Across Boundaries: Suppliers, Managed Providers, Sector Exercises and Authorities	Loops that cross organisational boundaries: managed providers and cloud, sector exercises, sharing without exposing weaknesses, disclosure and evidence to supervisors

Part VI: People and Career

The part is introduced by part_vi_people_and_career.md.

Chapter	File	Title	Description
23	23_red_and_blue_as_people.md	Red and Blue as People: Trust, Incentives, Safety and Wellbeing	The human side of an adversarial practice: incentives, blamelessness, being tested, wellbeing and staff exposed to exercises
24	24_competences_and_keeping_the_red_craft.md	Competences, Training and Keeping the Red Craft	Competences for the loop, competence records, the red apprenticeship problem; a constructive debate on whether automated emulation erodes the craft
25	25_the_future_of_the_loop.md	The Future of the Loop: Will Red and Blue Be Separate Teams in 2036?	Outlook under the three labels, observable, expected and speculation; a constructive debate on whether red and blue will be separate teams in 2036; the author's projections

Part VII: Experience and Conclusions

The part is introduced by part_vii_experience_and_conclusions.md.

Chapter	File	Title	Description
26	26_case_studies_and_anti_patterns.md	Case Studies and Anti-Patterns	Case studies and anti-patterns: purple theatre, the red team that grades its own homework, coverage by assertion, the green dashboard and the contract with no testing right
27	27_a_year_in_the_loop.md	A Year in the Loop: Nordvik Region and Hansa Mutual	A narrative year for four named people in the two organisations, following scenario, exercise, finding, change and re-test, including a failure and a recovery
28	28_conclusions_for_nordvik_and_hansa.md	Nordvik Region and Hansa Mutual: Conclusions for Two Organisations	A conclusion for each organisation and a comparison, with the plan for the next two years as a decision record
29	29_conclusion.md	Conclusions: The Purple Loop with AI	The author's verdict on the five-part thesis, the independence principle and the questions left open; the author's projections, labelled

Appendices and Reference

The appendices are introduced by part_appendices.md.

Item	File	Title	Description
Glossary	30_glossary.md	Glossary	Defines the vocabulary of the book.
Appendix A	31_appendix_a_code_examples.md	Appendix A: Code Examples: Scenario, Mandate, Coverage, Regression and Evidence Artefacts	Fuller versions of the artefacts introduced in the chapters.
Appendix B	appendix_templates_and_tools.md	Appendix B: Templates and Tools	A directory of the templates and tools the chapters provide, with the chapter that introduces each.
References	32_references.md	References and Sources	Canonical bibliography
About the Author	33_about_the_author.md	About the Author	Background and perspective

Companion files

File	Purpose
BOOK_BLUEPRINT.md	Premise, thesis, audience, the brief of every chapter, the foundations it builds on, the purple loop stages it concerns and its status
SERIES_FOUNDATIONS.md	The conclusions of Books 1–8 that this book builds on, with the tables of contents of the earlier volumes for cross-references
EXAMPLE_ORGANISATIONS.md	Canonical profiles of Nordvik Region and Hansa Mutual, including their purple profiles
CHAPTER_AUTHORING_GUIDE.md, STYLE_GUIDE.md, AUTHOR_PERSONA.md	Rules for structure, language and voice

Canonical ordering

scripts/toc_definition.py holds the table of contents and regenerates docs/book_index.json, the mkdocs.yml navigation and the chapter list in BOOK_REQUIREMENTS.md. docs/book_index.json and mkdocs.yml must agree (tests/test_book_index.py).