AI screening levels
Three levels of screening intelligence
From an instant deterministic Quick Check to AI-reviewed matches that flag likely false positives — choose the depth each case needs, across sanctions, PEP and criminal watchlists. EU-hosted.
Дмитрий Ковалёв → Dmitri Kovalev
Transliteration variant generated
OFAC SDN candidate
DOB mismatch: 1971 vs 1984
EU Consolidated List candidate
Name and nationality align
Reasoning + indicators
Exported to evidence PDF
The problem
More matches, or fewer false positives — most tools make you choose
AI screening applies machine intelligence to sanctions and AML name matching: it expands queries, resolves transliterations and aliases, and reasons about whether a hit is a real match. Instead of one rigid string comparison, ScreenVeritAI offers three levels — so teams catch more true matches while spending less time clearing noise.
Turn matching up and analysts drown in false positives; turn it down and a transliterated name slips through onto a live list. A single fixed algorithm cannot serve both a million-row monitoring run and a high-risk onboarding decision. ScreenVeritAI splits the problem into three levels you choose per check, so recall and reviewer time are tuned to the risk of each case rather than fixed for the whole account.
The three levels, side by side
Every level screens the same official sources. What changes is how names are matched, how much reasoning is applied, and what — if anything — is sent to AI models.
| Screening level | What it does | When to use it | Data sent to AI models | Price per check |
|---|---|---|---|---|
| Quick Check | Deterministic exact and structured matching across every source — no AI in the loop. | High-volume, lower-risk screening and mass monitoring where speed and cost matter most. | None — data never leaves the platform. | €0.79 |
| AI-Enhanced Check | Adds query expansion, transliteration and alias resolution to widen recall. | Non-Latin scripts, transliterated records, or anywhere a plain lookup may miss name variants. | Expanded name forms only, via EU-hosted models with zero retention. | €1.39 |
| AI Review Check | Adds AI-assisted reasoning and false-positive indicators for every candidate match. | Higher-risk onboarding, escalations and any hit you need to explain to an auditor. | Match context only, via EU-hosted models with zero retention. | €1.79 |
Prices are public pay-as-you-go rates per check. Volume discounts apply — see the volume table below. Figures as of June 2026.
Matching examples
Names a plain lookup misses — and AI-Enhanced catches
Real sanctions and PEP records rarely arrive in the exact form you type. These are the variant classes the AI-Enhanced Check resolves automatically.
A plain lookup misses this
Дмитрий Ковалёв
AI-Enhanced catches
Dmitri Kovalev · Dmitriy Kovalyov · Dmytro Kovalov
A plain lookup misses this
Abdul Rasheed
AI-Enhanced catches
Abd al-Rashid · Abdurrashid
A plain lookup misses this
William
AI-Enhanced catches
Bill
A plain lookup misses this
Eagle Pharmaceuticals
AI-Enhanced catches
Eagle Pharma OÜ
A plain lookup misses this
Łukasz Kowalski
AI-Enhanced catches
Lukasz Kowalski
A plain lookup misses this
Jürgen Müller
AI-Enhanced catches
Juergen Mueller
How the AI levels work
Each level builds on the one below it, so you add intelligence only where a case needs it.
Query expansion and transliteration
A plain lookup compares the exact string you typed against a list, and names rarely arrive that cleanly. The AI-Enhanced Check expands each query into the forms a real record might use: it romanises Cyrillic, Arabic and other scripts, generates spelling and diacritic variants, and accounts for word order and initials. A search for Дмитрий Ковалёв is checked against Dmitri Kovalev, Dmitriy Kovalyov and Dmytro Kovalov; Jürgen Müller is checked against Juergen Mueller. This expansion runs against the same official sources as the deterministic Quick Check, so you widen recall without changing your coverage. Legitimate matches hidden behind a transliteration or a stray accent surface as candidates instead of silently passing — the failure mode compliance teams fear most.
Alias, nickname and entity resolution
Sanctioned and politically exposed people appear under nicknames, married names, corporate suffixes and local-language forms. The AI-Enhanced Check resolves these: William is linked to Bill, Abdul Rasheed to Abd al-Rashid and Abdurrashid, and Eagle Pharmaceuticals to Eagle Pharma OÜ. Rather than making you think of every variant yourself, the platform generates the alias set and entity forms for you, then screens them together. For companies it also considers legal-form tokens — OÜ, GmbH, Ltd — that routinely break exact matching. Every candidate it returns is traceable back to the source record and the variant that triggered it, so an analyst can see why a name surfaced and confirm or dismiss it quickly rather than guessing at the tool's logic.
AI match reasoning and false-positive indicators
The AI Review Check adds a reasoning layer on top of the expanded search. For each candidate it compares the available identifiers — name, date of birth, nationality, entity type — and produces a short, readable rationale for why the record is or is not a likely match, along with explicit false-positive indicators such as a mismatched birth year or a different country. This is the level that attacks alert fatigue directly: instead of a queue of undifferentiated hits, your team sees which candidates are worth escalating and which carry clear reasons to clear. The reasoning and indicators are exported with the result, so the judgement is documented for your file. Use this level on higher-risk onboarding, escalations, and any hit an analyst needs to defend.
Explainable by design
Reasoning you can read and export — never a black-box score
AI Review never reduces a match to an opaque risk number. For every candidate it returns the reasoning behind the decision and the specific false-positive indicators it found — mismatched dates, different nationalities, weak name overlap — in plain language your analysts and your auditors can read, challenge and export with the result.
- A short written rationale for each candidate match
- Explicit false-positive indicators, not just a number
- Every candidate traceable to its source record and matched variant
- Reasoning and indicators included in the point-in-time evidence PDF
EU data residency
Deterministic modes send nothing to AI models
ScreenVeritAI is EU-hosted in Finland, and data at rest never leaves the EU. The deterministic Quick Check and continuous monitoring send no data to AI models at all. When you choose an AI-assisted level, screening runs on EU-hosted models with zero data retention.
- Quick Check and monitoring: zero data to AI models
- AI-Enhanced and AI Review: EU-hosted models, zero retention
- Data at rest stays in the EU (Finland)
- Mix levels per case — pay only for the intelligence you use
Volume pricing
Per-check rates fall as monthly volume rises. The same three levels are available at every tier.
| Monthly volume | Quick Check | AI-Enhanced Check | AI Review Check |
|---|---|---|---|
| Up to 500 / mo | €0.79 | €1.39 | €1.79 |
| 500+ / mo | €0.59 | €1.09 | €1.69 |
| 1,000+ / mo | €0.49 | €0.89 | €1.59 |
| 1,500+ / mo | €0.39 | €0.69 | €1.49 |
Public pay-as-you-go rates per check, EUR. Figures as of June 2026.
Pay per check — no platform fee
Choose a level per screening. There is no monthly subscription for screening intelligence.
Quick Check
Deterministic, no AI — data never leaves the platform
AI-Enhanced Check
Query expansion, transliteration and alias resolution
AI Review Check
AI-assisted reasoning and false-positive indicators
No monthly platform fee. Pay only for checks, reports and monitored entities.
ScreenVeritAI supports compliance review and due diligence documentation. Final decisions remain with your team.
Works together with
The AI levels run across every screening domain on the platform.