How It Works

From Entity Name to Audit-Ready Report in Under 60 Seconds

An 8-step deterministic pipeline — AI-powered reasoning, human review where it matters, and a cryptographically verifiable output your legal team can rely on.

Full Screening Pipeline

01
Input Entity

Submit the entity for screening

API + Dashboard

Enter a company name, individual name, or IDNO (Moldovan registry identifier). The system accepts free-text input, structured JSON via API, or bulk CSV uploads for portfolio screening.

02
Normalize

Deterministic name normalization

Deterministic Python · No AI

Zero LLM involvement at this stage. A deterministic Python pipeline applies Cyrillic→Latin transliteration (ICU standard), abbreviation expansion (LLC → Limited Liability Company, OOO → Общество), and diacritic stripping. This ensures reproducible, auditable normalization every time.

03
Direct Match

Cross-check against 7 live sanctions lists

OFAC · EU FSF · UK HMT · UN SC · SECO · OpenSanctions

Simultaneous exact + fuzzy matching against OFAC SDN, EU Consolidated Financial Sanctions, UK HMT, UN Security Council, SECO Switzerland, OpenSanctions, and national EU authority lists. Levenshtein distance with configurable threshold (default: 85%) catches transliteration variants.

04
UBO Analysis

Traverse the full ownership chain

ONRC · E-Gov Moldova · OpenCorporates

We resolve the beneficial ownership tree 3–5 levels deep using live ONRC Romania, E-Gov Moldova, and OpenCorporates APIs. Each entity in the chain is independently screened. Circular ownership detection and offshore obfuscation patterns are flagged automatically.

05
Indirect Exposure

Map commercial and banking relationships

Commercial exposure · Banking relationships

Beyond direct ownership: we identify sanctioned counterparties in the entity's supplier network, client base, and banking relationships. This catches secondary exposure — the entity isn't sanctioned, but it does business with one that is.

06
Adversarial Verify

AI challenges every RED finding

Multi-agent AI · False positive rate <5%

Our Adversarial Verifier agent actively searches for counter-arguments before any RED FLAG enters your report: documented divestiture, pre-sanctions ownership transfers, regulatory exemptions, OFAC licenses. Industry false positive rate: 15–25%. Ours: under 5%.

07
Human Review

Mandatory compliance checkpoint for RED findings

EU AI Act Art. 14 · Cannot be bypassed

All RED FLAG verdicts require a qualified human reviewer before a report can be generated. This is not optional — it is hardcoded compliance with EU AI Act Article 14 (Human Oversight). The reviewer identity, timestamp, and rationale are embedded in the final report.

08
Report

Audit-ready, legally defensible output

SHA-256 · RFC 3161 · PDF + JSON

The final report includes all 8 components (Executive Verdict, Direct Matches, UBO Tree, Indirect Exposure, Adversarial Log, Human Validation, Legal Disclaimer, Blockchain Timestamp). A SHA-256 hash and RFC 3161 timestamp provide immutable proof of report content at delivery time.

TECHNOLOGY

The Architecture Behind the Accuracy

Claude
Anthropic
Primary Reasoning
200K context · AWS Bedrock EU
Gemini
Google
Adversarial Verification
2M context · Vertex AI VPC
Grok
xAI
OSINT / Web Intel
Real-time web · API
ORCHESTRATOR AGENT
Session · Routing · State Management
Intake Agent
OCR · Classify
Graph Builder
Neo4j · Pinecone
Screening Agent
OFAC·EU·UK·UN·SECO
Adversarial Verifier
Claude + Gemini cross-check
Human Review
EU AI Act Art.14 enforced
Reporter
PDF · JSON · Blockchain hash
Anthropic ClaudeGoogle GeminixAI GrokAWS Bedrock EUNeo4jPineconePythonRFC 3161AES-256TLS 1.3

When Claude and Gemini reach the same conclusion on a RED FLAG finding, confidence is high. When they diverge — the divergence itself becomes the finding, escalated to human review. No single model's bias survives adversarial cross-verification.

What You Get: Audit-Ready, Legally Defensible Reports

Report Contents

01
Executive Verdict
CLEAR / YELLOW FLAG / RED FLAG with global risk score (0–100)
02
Direct Match Results
Source list, date of listing, official URL, unique ID
03
UBO Ownership Tree
Visual chain diagram + each entity sanctions status
04
Indirect Exposure Map
Suppliers / clients / banking relationships with RF connections
05
Adversarial Challenge Log
What we tried to disprove and why we kept or dropped each finding
06
Human Validation Record
Expert identity, timestamp, and override rationale (EU AI Act Art. 14)
07
Legal Disclaimer
Explicit AI result advisory; legal counsel recommendation
08
Blockchain Timestamp
RFC 3161 immutable proof of report content at delivery time

Verdict System

CLEAR

No direct matches. No sanctioned UBOs identified. No significant indirect exposure detected.

Proceed with standard commercial caution.

YELLOW FLAG

Indirect RF exposure identified. Inconclusive UBO chain. Fuzzy match requires manual review.

Proceed with enhanced due diligence. Legal review recommended.

RED FLAG

Direct sanctions list match OR sanctioned UBO confirmed. Critical risk identified.

DO NOT PROCEED pending legal counsel.

Ready to Screen Your First Entity?

Try the live demo with real sanctions data, or request a pilot for your compliance team.