An open map of every legal-data source in the Middle East and North Africa — gazettes, consolidated legislation, apex courts, sectoral regulators — with structured pointers for anyone contributing scrapers, indexes, or research on top of MENA open law.

Status: v1.0 — researcher-compiled with a URL-validation pass. Per-jurisdiction lawyer validation lands in v1.1.

mena-legal-atlas/
├── phase-0-audit/          # Audit of worldwidelaw/legal-sources MENA coverage
│   ├── mena-audit.md         Human-readable status table
│   ├── mena-audit.csv        Per-source rows
│   └── mena-audit.json       Machine-readable
├── phase-1-atlas/          # The Atlas itself
│   ├── atlas.yaml            Source-of-truth structured data (25 jurisdictions)
│   ├── mena-legal-atlas.md   Headline document
│   ├── mena-legal-atlas.json Machine-readable
│   └── priority-action-list.md  Ranked targets for Phase 2
├── scripts/
│   ├── audit_mena.py         Regenerates Phase 0 from upstream manifest
│   └── generate_atlas.py     Compiles Phase 1 outputs from atlas.yaml + audit
└── data/legal-sources-upstream/   # shallow clone of worldwidelaw/legal-sources

• 118 MENA sources cataloged in worldwidelaw/legal-sources across 24 jurisdictions.
• Only 36% currently produce data (42 working, 76 blocked).
• 10 jurisdictions have zero working scrapers: Lebanon, Saudi Arabia, Israel, Iraq, Iran, Palestine, Sudan, Somalia, Comoros, Yemen.
• Yemen has no entries upstream at all — the only completely greenfield MENA jurisdiction.
• The dominant blockers are Cloudflare WAF, geo-IP restrictions, third-party paywalls, and login walls — not missing source code.

Top 5 priority targets (composite score = ICP × 3 + upstream gap × 2 + tractability bonus):

See phase-1-atlas/priority-action-list.md for the full ranking.

The atlas maps where the law lives. The scorecard scores where you can actually use it. Across the 25 jurisdictions: 20/25 have full 5-category source depth, but only 4 are actually reachable (4+ live sources) and 10 are fully blocked (0 live sources). Mapped is not the same as reachable, current, or in your language.

Five axes per jurisdiction (open viz/coverage-scorecard.html):

• Source depth — how many of the 5 source categories exist (local, from the atlas).
• Reachability — working-source tier (hi / mid / lo), not raw %, to avoid the small-denominator trap (1 working / 0 blocked ≠ "100% reachable").
• Language regime — Arabic-only vs bilingual vs other (the HAQQ wedge).
• Recency — any major 2025–26 reform, checked via Linkup constrained to each jurisdiction's official domains; every "active" flag carries its source link and is only marked yes when the sources support it.

python3 scripts/score_coverage.py     # -> data/coverage-scores.json (needs Linkup + OpenRouter keys)
python3 viz/build_scorecard.py        # -> viz/coverage-scorecard.html

See the live-data comparison in phase-2-enrichment/BENCHMARK-vs-legal-data-hunter.md (this atlas vs Legal Data Hunter, recomputed from LDH's public status.json).

Open phase-1-atlas/mena-legal-atlas.md — it's the headline document.

Consume phase-1-atlas/mena-legal-atlas.json — it's the machine-readable version with reconciled upstream source IDs.

Atlas entries are in phase-1-atlas/atlas.yaml. Open a PR with edits. Run python scripts/generate_atlas.py to regenerate downstream files.

cd data/legal-sources-upstream && git pull
cd ../.. && python scripts/audit_mena.py && python scripts/generate_atlas.py

25 jurisdictions:

• Arab League (22): DZ, BH, KM, DJ, EG, IQ, JO, KW, LB, LY, MR, MA, OM, PS, QA, SA, SO, SD, SY, TN, AE, YE
• Non-Arab MENA (3): IR (Iran), IL (Israel), TR (Turkey)

Edge cases are flagged in each country entry. The Atlas takes no position on contested sovereignty (Iraqi KRG vs Federal, Libya GNU vs HoR, Yemen IRG vs Houthi, Palestine PA vs Hamas, post-2024 Syria) — it documents the parallel sources where they exist.

Every entry is currently unverified. v0.1 is researcher-compiled from public sources, the upstream manifest.yaml, and the maintainer's domain knowledge. Do not treat as authoritative.

Phase 2 of the project will pair each jurisdiction with a local lawyer or law-school validator. Validator credit will be added per entry.

• Atlas content (this repo): CC BY 4.0 — use, modify, redistribute with attribution.
• Upstream worldwidelaw/legal-sources is AGPL-3.0 with a commercial license. We do not bundle their scraper code — only cross-reference their source IDs. Any contributions back to upstream will be governed by their CLA.

HAQQ Legal AI is a MENA-focused legal AI company. Operating in the region quickly surfaces a fact that no global index captures: every country publishes its laws somewhere, but the patchwork of gazette PDFs, academic mirrors, regulator rulebooks, and court portals has never been documented in one place.

The Atlas is the map. The corpus is the byproduct.

Two new layers were added to atlas.yaml via a sourced-search API (Linkup), staged and filtered before merge — see linkup-smoke-test.md and phase-2-enrichment/PRECISION-BENCHMARK.md:

• additional_sources — 233 official-source domains (central banks, tax/data-protection/ competition/IP authorities, courts, registrars) the gazette-centric map missed. Domain-level discovery is where AI search reliably works.
• key_legislation — constitutions for 21 jurisdictions (via Constitute Project / WIPO Lex)
  • 1 verified primary entry (Lebanon).

Honest caveat: building a document-level laws catalog by AI search caps at ~3% precision — MENA portals are search-DB SPAs whose law text no crawler indexes. The benchmark documents why.

Two-panel choropleth (regenerate via viz/build_map.py):

• Where the law lives — source coverage, ~80% of jurisdictions complete across 5 categories.
• Where you can reach it — only 36% of catalogued sources produce machine-readable data.

The gap between the two panels is the core finding: in MENA, mapping the law is largely solved; retrieving it is not.

phase-2-enrichment/BENCHMARK-vs-legal-data-hunter.md compares this atlas against Legal Data Hunter (the global crawler / worldwidelaw/legal-sources, which this atlas is built on), using LDH's live public status.json:

• LDH is 56% complete globally but only 41% across MENA (163 sources) — MENA is its weakest slice.
• This atlas surfaced 152 official MENA source domains LDH does not yet track — direct candidates for upstream collection scripts.

They're complementary: Legal Data Hunter retrieves; this atlas maps where MENA's law lives and feeds that map back upstream.

• v1.0 (now) — researcher-compiled Atlas + Phase 0 audit + URL validation
• v1.1 — per-jurisdiction validation by local lawyers (starting LB + SA)
• v1.2 — interactive web view
• v2.0 — scraper contributions land upstream for top 5 priority jurisdictions

Maintained by Stephane Boghossian (Head of Growth, HAQQ Legal AI). Atlas authorship is a HAQQ-sponsored open project; opinions are the maintainer's.