The Global PEP Census

Data as of 2026-08-11 (corpus build) · PEP corpus last expanded 2026-07-11–12

No sanctions-data aggregator — including OpenSanctions, the open-data reference implementation this field measures itself against — publishes a politically-exposed-persons dataset sourced entirely from official government and international-organization publications, at this scale. We do. As of this writing, our corpus includes 51,552 PEP records spanning all 199 jurisdictions (193 UN member states, 2 UN observer states, and 4 non-UN jurisdictions we also cover: Hong Kong, Taiwan, the Cayman Islands, and Kosovo) — and as of 2026-07-11, it contains zero records sourced from Wikidata.

That second fact is a deliberate trade, not a free upgrade, and we want to be upfront about what it cost. Wikidata is crowdsourced. It is fast to query and broad in reach, but a compliance team building an audit trail should be able to answer "where did this fact originate" with an institution's own publication, not an editable wiki. We removed our Wikidata-sourced PEP layer (wd_peps) from the published corpus in favor of citing each record to the issuing government or organization's own page, API, or bulk export — a parliament's members directory, a cabinet page, a central bank's leadership page, a CIA World Leaders country profile. The removal cost real breadth: depth in judiciary, military, and central-bank roles, and coverage of formerly-in-office officials predating August 2022, is measurably thinner than it would be with Wikidata still in the mix. We made that trade anyway, because a PEP flag exists to support a regulated institution's due-diligence decision, and every record in this census can be traced back to the government or institution that made the underlying fact true — the standard we'd want if a regulator asked "where did this come from."

What we counted, and how

A politically exposed person (PEP) is not a sanctioned party. Under FATF Recommendation 12, a PEP flag means only that someone holds — or recently held — a prominent public function, which warrants enhanced due diligence, not automatic refusal. We classify every record in this census along two independent axes:

• Government branch (B1–B9): Executive (B1), Legislature (B2), Judiciary — apex/constitutional courts only (B3), Military/Security leadership (B4), Central bank/monetary authority (B5), State-owned enterprises (B6), International-organization officials (B7 — FATF's "international PEP," a role-based overlay that applies regardless of home jurisdiction), Subnational/regional government (B8), and Political-party senior leadership (B9).
• Seniority Level (1 / 2 / 3 / RCA): Level 1 is apex state power — heads of state and government, cabinet ministers, apex judiciary, top military/security chiefs, central bank governors, and heads of international organizations. Level 2 is senior officials one step below apex — national legislators, ambassadors, deputy ministers, SOE leadership. Level 3 is subnational and lower-visibility senior roles. RCA is relatives and close associates of a Level 1 or 2 PEP, tracked as a relationship, not a role.

Every record also carries a current/former status computed from term dates, and a former PEP's classification decays on a published, Level-driven retention window rather than persisting indefinitely: Level 1 retains for 50 years, Level 2 for 20 years, Level 3 for 5 years, and RCA inherits its anchor PEP's window (these figures are adopted from OpenSanctions' published methodology, mapped onto our four Levels — we did not invent our own retention schedule).

Coverage by tier

We reach all 199 jurisdictions through three sourcing tiers, and every jurisdiction — regardless of tier — contains positions spanning the full Level 1–RCA range; a tier-3 country's head of state is still Level 1.

Six tier-2 jurisdictions (Brazil, Canada, Italy, Norway, Poland, Sweden) and all five tier-1 sovereign jurisdictions already have a live national-legislature dataset, ahead of their full hand-sourced build-out schedule — a branch can ship early without waiting on its tier's complete depth plan.

The 17 live datasets

Per-dataset counts are engineering's vendored dataset sizes at ship time; the published total of 51,552 reflects final cross-source dedup, so the sum of this column runs slightly higher than the published total — expected and disclosed, not an error.

By seniority Level

Level 3 is live for the first time with this expansion, and it is carried entirely by the US state/subnational dataset — it is not yet global. Expanding Level 3 beyond the US is roadmap work (EU first, tracking the EU AMLR's 2027 subnational reporting obligation), not a claim we make today.

Date-of-birth coverage — shown, not assumed

A name match means little without a date of birth to disambiguate it. We report DOB coverage the same way we report everything else: labeled by confidence, not blended into a single number.

• 98.5% of live PEP profiles display a date of birth or a dated range.
• 48.3% (24,890 profiles) carry a native, source-published date of birth, rendered as unlabeled sourced fact. This is concentrated in the newer legislature/subnational datasets — several publish complete birth records; others, including the UK and Australian parliaments, publish none.
• The remaining coverage comes from a labeled DOB-estimate layer, and every one of these renders "Estimated — not verified" with its method, evidence, and confidence shown alongside — never blended with the native figure:
  • 262 dates propagated from a reconciled same-person record elsewhere in the corpus.
  • 235 dates derived from a secondary source carrying a quoted evidence passage and link (tiered official > news > crowdsourced; a crowdsourced or Wikipedia-derived lead is explicitly flagged as a lead, not an authoritative source).
  • ~25,376 an office minimum-age bound ("born on or before <date>") inferred from the position held.
• 1.5% (789 profiles) carry neither a native DOB nor an estimated bound.

The purpose of the estimate layer is disambiguation — ruling a name match in or out — not proof of identity.

What's not yet covered

We would rather tell you what's missing than imply completeness we haven't earned:

• National legislatures are live for 13 jurisdictions (UK, Canada, Australia, Germany, France, Brazil, Sweden, Norway, Netherlands, Poland, Italy, the EU, and the US). The remaining jurisdictions' legislatures are on the roadmap, not yet sourced.
• Subnational government is live only for the United States. Expansion beyond the US is planned but not shipped.
• Political-party senior leadership (Branch B9) is not yet sourced anywhere in the corpus.
• Branch B7 (international organizations) is live for 21 of roughly 38 scoped organizations, at apex/deputy/board level — not full staff directories. A long tail of UN specialized agencies, political/security blocs, and financial-standard bodies remains on the roadmap.
• Judiciary, military, and central-bank depth beyond the CIA World Leaders apex baseline is not yet built out for most tier-1/tier-2 jurisdictions.
• RCA (relatives and close associates) is not yet indexed anywhere in the corpus.

None of these gaps are backfilled by re-adding Wikidata. Every gap above will close, if it closes, through another official-source, primary-publisher dataset.