Append-only. Newest entries at the bottom.
2026-07-28 — Initial ingest
Sources ingested: raw/Laws/ — 16 files, ~152,200 words.
| Source | Words | Coverage |
|---|---|---|
ead_v2.md (IEEE EAD v2) | 94,100 | Chapter/section structure + front matter only |
NIST.AI.100-1.md (AI RMF 1.0) | 14,500 | Exec summary, §3 characteristics, §5 functions, Appendix B — read in full |
Overview of the Code of Practice...md | 7,600 | Exec summary, Safety & Security chapter, Appendices 1 and 4 — read in full |
What Is AI Governance_Paloaltonenetworks.md | 7,700 | Understanding, challenges, audits, IR plan, Securing AI Systems — read in full |
OECD_Legal_Instruments_OECD.md | 4,100 | Sections 1 and 2 — read in full |
Modifying AI Under the EU AI Act...md | 4,100 | Summary + headings |
Whistleblowing and the EU AI Act...md | 3,050 | Summary + headings |
Enforcement of Chapter V...md | 2,700 | Summary + headings |
Providers of GPAI Models...md | 2,450 | Headings |
The EU AI Act's Transparency Rules (Art 50)...md | 2,200 | Summary + headings |
What the EU AI Act Means for Staffing Businesses...md | 2,260 | Summary + headings |
Singapore’s_Approach_to_AI_Governance_PDPC.md | 1,900 | Read in full |
Overview of Guidelines for GPAI Models...md | 1,800 | Headings |
AI Regulatory Sandbox Approaches...md | 1,375 | Read in full |
The Scientific Panel...md | 1,300 | Summary + overview |
The Advisory Forum...md | 854 | Summary + overview |
Coverage caveat: IEEE EAD v2 is 62% of this collection’s word count and was read only structurally. Its wiki page reflects that and says so.
Pages created (14):
index.md,log.mdnist-ai-rmf.md,trustworthy-ai-characteristics.mdoecd-ai-principles.md,singapore-model-framework.md,ieee-ethically-aligned-design.mdgpai-code-of-practice.md,code-of-practice-security-mitigations.md,systemic-risk.mdai-governance-practice.mdeu-ai-act-governance-bodies.md,ai-regulatory-sandboxes.md,whistleblowing.mdhigh-risk-ai-in-employment.md
Note on placement: several EU AI Act explainers physically in raw/Laws/ interpret specific articles and are cited primarily from ai-regulations-wiki pages (Art 50 guide, Chapter V enforcement, Modifying AI, staffing businesses, GPAI guidelines). Their governance-side content lives here. Cross-links run both ways.
Findings recorded:
- Code of Practice Appendix 4 is the corpus’s only named-control security baseline; it defends weight confidentiality, not the injection/agentic surface.
- The systemic-risk capability/propensity/affordance split maps cleanly onto three different mitigation surfaces; most agentic risk is affordance, which the integrator controls rather than the model provider.
- Singapore is the only framework here that ships testing tooling (AI Verify).
- ENISA holds a permanent Advisory Forum seat; the Scientific Panel can trigger systemic-risk designation without the 10²⁵ FLOP threshold.
- Cryptographic signing of inbound models/datasets appears only in the vendor source, not in Appendix 4.
Root index updated: yes.
Open items:
- IEEE EAD v2: only the v2 public-discussion draft is present. The final EAD1e (2019) should be sourced before citing anything as current IEEE guidance.
- NIST AI RMF Playbook and the AI RMF Generative AI Profile are not in
raw/. - The Code of Practice source is a third-party overview, not the Code itself.
- No source covers governance of agentic deployments specifically.
2026-07-28 — Lint fix and ATLAS cross-linking
Defect fixed: eu-ai-act-governance-bodies.md did not record a contradiction between two sources, violating the CLAUDE.md citation rule that disagreements be noted explicitly.
The Scientific Panel source calls the Panel “one of the three advisory bodies, alongside the AI Board and Advisory Forum”; the Advisory Forum source calls the Forum “one of the two advisory bodies, in addition to the Scientific Panel.” Both are from the same publisher. The discrepancy turns on whether the AI Board (Art 65) counts as advisory. Now recorded on the page with that resolution.
Stale claims corrected: ai-governance-practice.md and index.md both stated raw/MITRE-ATLAS/ was empty. ATLAS is now ingested — see mitre-atlas-wiki/. Both updated to point at it, including the technique-level mitigations that close the gap index.md flagged.
2026-07-30 — Collection renamed ai-governance-wiki → laws-wiki
Request: “change the AI-governance wiki name to Laws wiki.”
Context: the curator had renamed raw/AI-governance/ → raw/Laws/. Earlier the same day, 61 source-path citations across 40 wiki files were repaired to match. This rename brings the wiki folder into line with the source folder, restoring the raw/<X>/ → wiki/<x>-wiki/ pairing convention for this collection.
Changes:
- Folder
wiki/ai-governance-wiki/→wiki/laws-wiki/(15 pages, contents unchanged). - 83 path references to
ai-governance-wikirewritten tolaws-wikiacross 49 files — every collection in the vault links here, so this touched more files than any previous single change. - Titles:
# AI Governance — Index→# Laws — Index; this log’s heading likewise. - Collection display labels in wikilinks:
|AI Governance]]→|Laws]](4 instances),|AI Governance collection]]→|Laws collection]](1). Page-level labels were deliberately left alone —[[../laws-wiki/nist-ai-rmf|NIST AI RMF]]and the CoP Appendix 4 links name specific documents, not the collection, and renaming them would have damaged 70+ accurate citations. One label,|the governance collection]], was also kept: it reads as prose and remains true. README.md: collection table row relabelled, and the source-pairing note rewritten now that the convention holds.
Name versus content. The folder now says laws, and the content is still predominantly governance frameworks — NIST AI RMF, OECD AI Principles, Singapore’s Model Framework, IEEE Ethically Aligned Design — which are voluntary instruments, not statute. The EU AI Act’s institutional machinery (Advisory Forum, Scientific Panel, governance bodies) sits here too, while the Act’s substantive obligations live in ai-regulations-wiki. Read laws-wiki as a pointer to raw/Laws/, not as a description of the subject matter. Recorded in the index’s Gaps section so a future reader does not mistake the folder name for a scope claim.
Unrelated defect found and fixed during verification: singapore-model-framework.md and this log both cited the Singapore PDPC source with an ASCII apostrophe (') in the filename, while the file on disk uses a curly apostrophe (U+2019, ’). Visually identical, dead as a path, and it survived every manual review until a scripted check caught it — almost certainly copy-paste normalisation. Both occurrences corrected to the exact on-disk filename. (The bad path is described here rather than quoted, so this log entry does not itself register as a dead citation in the source-path checker.)
Verification: zero remaining references to ai-governance-wiki anywhere; all 1,401+ wikilinks re-checked (5 pre-existing false positives in ATLAS pages only); every full-path raw/Laws/... citation now resolves against disk.
Known limitation, not fixed: many citations to this collection’s sources use ellipsis-abbreviated filenames — e.g. `raw/Laws/Overview of the Code of Practice...md` — because the real filenames run to 70+ characters. Readable, but not machine-checkable, so an automated integrity check cannot verify them. Left as-is rather than rewriting ~16 citations across 20 files; flagged here so the limitation is known.