README

AI and Data — Wiki Vault

Owner: Paul Miles (Red Strata) Domain: AI and data governance / assurance in EU & UK financial services Seeded: 2026-05-17 from Open Brain corpus (339 thoughts, 15 Mar – 13 May 2026)

Orientation for any AI agent operating on this vault

1. Read wiki-schema in /_schema/ first — it is more authoritative than any individual prompt.
2. Every claim on a synthesis page ends in a [S-tag] referring to a Source page in /sources/.
3. Never invent examples. Never silently resolve contradictions — surface them in ## Tensions or a Debate page.
4. Raw source material (the Open Brain dump for this seed) lives in /_raw_sources/ and is never edited.

What this wiki covers

The wiki was seeded from a daily regulatory-intelligence capture routine plus active client engagements. The synthesis surfaces are:

• EU and UK AI regulation — EU AI Act, FCA “no separate rulebook” approach, BCBS / BIS supervisory direction, EBA Work Programme.
• Operational resilience and third-party risk — DORA, FCA PS26/2, BCBS d605, Critical Third Parties regime.
• Model risk management — SR 11-7 / SS1/23 scaling to GenAI and agentic AI, model inventory expectations.
• Data governance substrate — BCBS 239, data lineage, data ownership, IRB data lifecycle.
• Practitioner governance maturity — recurring CDO/CRO survey signals, AGILE framework.
• Active client work — CLM Transformation Programme at an Irish Tier-1 bank, IRB briefing, CLM Pilot.
• Practice positioning — Independent Specialist (Data & AI Governance), four service lines, Big-4 differentiation.

Folder layout

/_schema/            wiki-schema.md (standing instructions)
/_raw_sources/       untouched ingested sources, named by S-tag
/sources/            one Source page per S-tag
/topics/             Topic pages (synthesis)
/entities/
  /companies/        regulators, vendors, competitors
  /technologies/     regulations, frameworks, tools
  /concepts/         theoretical / methodological ideas
/projects/           active and archived project pages
/maps/               index / map pages
open-questions.md    global aggregator
README.md            this file

Seeding session notes

The initial seed (2026-05-17) processed 339 Open Brain thoughts. Heavy duplication was collapsed: BIS othp90 appeared in ~15 near-identical thoughts, the FCA Mills Review in ~10, FIFAI II / AGILE in ~8. Each was consolidated into a single Source page; the duplicates are recorded as ingestion artefacts rather than fresh sources. Six off-topic items were explicitly excluded by Paul: home AI server / Mac Mini stack, Sea Kayak project, subscription cancellations, Claude account mismatch, Open Brain workflow note, and the Data Management AI Knowledge Community venture concept.

A follow-on session is needed to write Source pages for every distinct external source individually (~35 in total) — the seed prioritised Topic, Entity, and Project pages first.