The perimeter, without rounding
Four perimeters, four different answers. The figures below come from the dated local index; they are not a global count of sanctions.
| Perimeter | Designations | African entries | Crypto addresses |
|---|---|---|---|
| OFAC SDN | 19,416 | — | 505 |
| UN Security Council | 1,010 | — | 0 |
| European Union | 0 | 0 | 0 |
| OpenSanctions Africa | 10,175 | — | 1 |
The European Union perimeter is not loaded in this instance. It is shown at zero, not estimated. The OpenSanctions Africa count covers five national lists and is not merged into the OFAC/UN designation count.
Piste 1 — widening the sources
OpenSanctions / yente: what it fills, what it displaces. The collection now covers 95 aggregated sanctions sources and 2,888 published wallets.
What it does not fill: the African crypto link moves from 0 to 1 wallet out of 2,888, or 0.03 %. The gap is structural, not a collection defect.
yente (a self-hostable engine) remains the right building block to query several lists without rebuilding name matching. The public API returns 401 without a key: no request is made from this page.
The five national African lists
Five published datasets, 10,175 entities in total. None of them publishes a crypto address.
| List | Country | Entities | Targets |
|---|---|---|---|
| Egypt Domestic Terrorist List | Egypt | 7,687 | 3,618 |
| South Africa Targeted Financial Sanctions | South Africa | 1,992 | 991 |
| Tunisia National Counter-Terrorism List | Tunisia | 332 | 166 |
| Nigeria Sanctions List | Nigeria | 138 | 69 |
| Kenya FRC Domestic Terrorism List | Kenya | 26 | 13 |
Their interest is not a flow proof: they provide names and countries that OFAC and the UN do not cover, therefore material for a future reference dataset. A name is not a wallet.
Pistes 2, 3 and 4 — from observation to a reference dataset
Three ways to rebuild the link that the lists do not publish.
Piste 2 — cross-check the blockchain
Start from a designated person, look for their addresses on-chain, then follow the flows. The A-01 tracer already shows the movements of an address; it does not attribute an identity. This step requires chain-analysis data, not a deduction from a name.
- Done: reading the 505 published addresses and their assets.
- To do: address → counterparty → entity graph, with source and date per edge.
- Limit: a shared address does not designate a person.
Piste 3 — the human circuit
The information missing from the lists circulates elsewhere: platform KYC, financial intelligence units, regional mechanisms (GIABA in West Africa), suspicious transaction reports. Documenting that circuit is a public-policy recommendation, not code.
- Done: the limit is named and linked to the source.
- To do: map contact points by country.
- Limit: no personal data is collected here.
Piste 4 — build the reference dataset
The real product is not one more list: it is a dated reference dataset linking person ↔ company ↔ address ↔ source, with a country/regime codebook (54 + 1), a transliteration engine, and a SHA-256 fingerprint per snapshot. The building blocks already exist: A-05 screening, A-04 BI, A-01 tracer. What is missing is the link between them — and the temporal version.
Horizons — what is done, what remains a perspective
No announced function is presented as available if it is not.
| Horizon | Status | What it means |
|---|---|---|
| 1 · Living observatory monitoring, alerts, time series | partial | Only one snapshot is kept (bi_history.json = 1). Daily monitoring, alerts and entry/exit detection between two dates are not implemented. |
| 2 · Relationship graph address ↔ entity ↔ flow | partial | The tracer reads one address and its movements. The multi-hop graph exists in A-08; the entity graph and circumvention paths are not implemented. |
| 3 · AI that reasons transliteration, score, assistant | deterministic | Arabic/Cyrillic/Greek transliteration, onomastic variants and similarity scoring are deterministic and explainable. No LLM, no opaque score, no natural-language assistant in this version. |
| 4 · Policy and academic impact publication, recommendation | published | The “zero African address” finding is published and recalculable. The recommendation to regulators (GIABA, central banks, CERTs) remains to be formalised in a dedicated note. |
| 5 · Product / service API, SaaS, subscription | not done | No public API, no account, no stored data: the site is static. An “Africa-first” screening service is a perspective, not an active offer. |
What to retain
Sources: OFAC SDN, UN Security Council, OpenSanctions (five African national lists). Snapshot 18 October 2026; review 20 October 2026. The French A-06 page contains the live counters and the monitoring log. A-05 and A-06 consoles are currently French-language.