AI / LLM Index
Topic TypeIndex PagePrimary UseSafe loading order for LLMs, RAG systems and local AML agentsScopeJurisdiction-neutral public referenceLast ReviewedJuly 2026
This index helps an AI system retrieve the Wiki as a controlled knowledge layer. It does not grant the model decision authority.
Recommended loading order#
1. Editorial & Source Policy
2. Source Register
3. Why AML Exists
4. AML Learning Path
5. Core Concepts
6. CDD / EDD / Beneficial Ownership / SoW-SoF
7. PEP / Sanctions / Adverse Media
8. AML Programme Design & Effectiveness
9. Sector Risk Lens and defensive typologies
10. Operation workflows and evidence lifecycle
11. MLRO escalation and authority boundaries
12. TCAML verification, S010 and orchestration controls
Retrieval by task#
| Task | Retrieve |
|---|---|
| Explain a concept | Canonical topic page + glossary |
| Build a review plan | Topic page + workflow + evidence lifecycle |
| Assess SoW/SoF evidence | SoW/SoF + EDD + sector page |
| Review a typology indicator | Typology + sector lens + source policy |
| Prepare escalation | Discrepancy controls + MLRO review + applicable overlay |
| Design a programme control | Programme Design + governance + QA |
| Configure a local agent | Operation workflow + verification control + S010 + RAG guide |
Agent boundary#
The system may collect, compare, summarise, draft, identify discrepancies and prepare review-ready outputs. It must not:
- invent missing facts or sources;
- claim live screening without an actual approved source check;
- convert red flags into allegations;
- conceal uncertainty;
- auto-confirm a stagegate;
- make final legal, sanctions, reporting, approval or exit decisions.
Canonical-page rule#
Where multiple pages mention a concept, retrieve the canonical page from the Topic Map. This reduces duplicated rules and contradictory fragments.