01
Donations register
Declared donors from the >$20k donations register, linked to the office-holders they gave to.
Solutions
·
For AML reporting entities
NZ-domestic PEP & close-associate screening.
Screen names against New Zealand’s public-office holders — ministers and MPs, the judiciary from the Supreme Court to the District Court bench, defence leadership, heads of mission, the Reserve Bank, Public Service chief executives, and the boards of state-owned enterprises — with per-record public-source provenance, 12-month lapse tracking, and an evidence-backed close-associates graph.
GET /v1/screening/coverage · live
| Dataset | Live records | Last verified |
|---|---|---|
| companies-officers | 161 | 10 Sep 2026 |
| district-court | 212 | 7 Sep 2026 |
| gg | 1 | 2 Sep 2026 |
| judiciary | 66 | 7 Sep 2026 |
| mfat-hom | 58 | 7 Sep 2026 |
| nzdf | 5 | 2 Sep 2026 |
| parliament | 172 | — |
| psc-ce | 41 | 7 Sep 2026 |
| rbnz | 11 | 7 Sep 2026 |
| treasury-soe | 39 orgs tagged | 2 Sep 2026 |
| close-associate edges | 2787 |
How it works
Ranked candidates, point-in-time answers, and evidence on every edge.
→
Name-only matching, ranked candidates.
Never a yes/no — exact, strong, and fuzzy tiers with scores. Your reviewer decides.
→
Point-in-time answers.
Holdings are append-only; as_at reproduces what the dataset said on any past date — evidence for your own audit trail.
→
12-month lapse tracking.
A holder who leaves office stays flagged as lapsed for 12 months rather than vanishing from results.
→
Delta feed for ongoing CDD.
Poll /changes for holdings opened or closed, lapse transitions, and associate edges — no re-screening.
→
Explainable associates.
Every edge carries the underlying public records — donation, directorship, declared interest, diary meetings. No black-box flags.
→
Batch, API, and MCP.
500 names per call, ~10 ms per name amortised; single checks 111 ms p50 / 133 ms p95 measured. Agent tooling speaks the same gated interface.
The associates graph
Built from records you can check, not signals you have to trust.
02
Directorships & interests
Companies Register directorships and declared pecuniary interests — shared boards and business relationships, with the register record on every edge.
03
Ministerial diaries
Repeated declared meetings from published ministerial diaries — who actually has access, on the public record.
What this is — and is not
A data feed for risk-based screening. Not a sanctions list, not a verdict.
→
Compiled exclusively from public sources.
The source URL is recorded on every record. Inclusion in the dataset is not an allegation of wrongdoing.
→
Name-only matching — human review required.
False positives are expected and results require human review. The Open Brief provides information, not legal or compliance advice.
→
Not a sanctions list, not identity verification.
A data feed for AML platforms and reporting entities doing risk-based screening of NZ persons — nothing more.
→
Not browsable on the public site.
Access is API-only, under an organisation-level agreement, with every screening call audit-logged.
Measured coverage
Benchmarked against the alternatives — and reproducible.
In July 2026 we enumerated the full register through this API — 685 office holders in scope across seven strata (current holders plus everyone whose office lapsed within the trailing 12 months), including the entire District Court bench. We then screened a deterministic stratified sample against the paid tiers of two commercial screening APIs, and the full census against the OpenSanctions open-data PEPs file.
depthcheck — found-at-all rate by stratum · 16 Jul 2026 · seed 20260716
| Stratum | OpenBrief | OpenSanctions | Vendor A | Vendor B |
|---|---|---|---|---|
| MPs & ministers (control) | 100% | 100% (125/125) | 100% (10/10) | 90% (9/10) |
| Senior judiciary | 100% | 15% (10/68) | 100% (8/8) | 38% (3/8) |
| District Court judges | 100% | 0% (1/209) | 87% (13/15) | 13% (2/15) |
| SOE / Crown boards | 100% | 3% (4/149) | 13% (2/15) | 80% (12/15) |
| Senior public service | 100% | 5% (2/41) | 60% (6/10) | 40% (4/10) |
| Diplomatic / defence / RBNZ | 100% | 15% (11/74) | 50% (5/10) | 80% (8/10) |
| Lapsed (last 12 months) | 100% | 0% (0/19) | 20% (2/10) | 90% (9/10) |
| All strata | 100% (685/685) | 22% (153/685) | 59% (46/78) | 60% (47/78) |
→
The databases are Parliament-shaped.
Every source finds MPs — that row is the control. Below Parliament, the two strongest commercial sources each miss about 4 in 10 office holders, and they miss different ones. A reporting entity would need both — and still have gaps.
→
One honorific turns a hit into a false clear.
Put “Hon …” in front of a name and both commercial vendors return a clean no-match — on people they actually hold. A false clear is worse than a miss. The register passed all 70 honorific and macron probes.
→
Nobody serves the close-associate obligation.
Close-associate screening is a statutory AML/CFT requirement. The register carries 2,506 evidence-backed edges — shared directorships, donations, declared meetings. No other source in this comparison links even one. Business-relationship screening is effectively unserved.
→
Depth, dates, and defensibility.
Finding a name is not the job — defending the decision is. The best comparator identifies the right office 74% of the time; the register always does, and cites the public record for every holding, with a last-verified date on each.
→
Point-in-time is unique in this comparison.
“Prove what you knew at the date of the transaction.” The register answers as at any past date — 67/67 point-in-time probes correct, every lapsed holder dated. Nothing else in this comparison can do it.
Notes on method
- Benchmark run 16 July 2026 on the production dataset; commercial vendors screened on their paid PEP-flagship tiers, not trials.
- Every number traces to an archived, timestamped run manifest; pilot customers can reproduce the census from seed 20260716 using the same API they integrate.
- The benchmark is frozen at its run date — the dataset panel at the top of this page is live, so the register will have moved since.
- Vendor stratum-level percentages carry small per-cell samples (8–15 names); the full report prints n in every cell, and the ALL row is the headline figure.
- Vendor B’s perturbation probes used a smaller sample than Vendor A’s; its honorific result (0 hits) is directionally consistent with Vendor A’s 0/30.
- Open-data comparison: OpenSanctions PEPs dataset, 15 July 2026, CC-BY-NC — © OpenSanctions, used under licence for evaluation; the bulk data is not redistributed.
- Named vendor results, per-cell samples, and raw run manifests available to prospective customers on request.
- Full methodology, manifest hashes, and reproduction commands: Methodology.
Pricing
Pilot pricing, sized to your book.
Practice
$400
/mo pilot · converts to $3,600/yr
Law firms, REAs, boutique trustees.
- 1,000 names monitored
- 250 checks / month
- Weekly changes feed
- Email support
Pilot target
Book
$750
/mo pilot · converts to $9,000/yr
Fund managers, fintechs, trust cos.
- 10,000 names monitored
- 2,500 checks / month + batch
- Daily changes feed
- Priority support + MCP
Portfolio
$1,500
/mo pilot · converts to $24,000/yr
Casinos, KiwiSaver, large books.
- 100,000 names monitored
- Unmetered checks (fair use)
- Webhooks, real-time changes feed
- SLA 99.9%, named contact
Channel / OEM licence
AML platforms and IDV vendors reselling the register — flat data licence plus per-end-client wholesale rate.
From $30,000
/yr · custom
Access
Want to try a pilot? Get in touch.
Onboarding is by agreement — tell us your reporting-entity or platform context and we’ll set up an API key.