SAYARI COMMERCIAL WORLD MODEL
The data substrate that enables judgment over AI.
A continuously enriched, self-repairing model of corporate and trade networks comprising three layers – Source Documents, a Knowledge Graph, and an Ontology.
world_model.resolve
{
“query”: “Meridian Trading LLC”,
“entities_resolved”: 4,
“jurisdictions”: [
“US-DE”, “CY”, “VG”
],
“ownership_depth”: 3,
“beneficial_owner”: “PEP_MATCH”,
“trade_records”: 847,
“risk_signals”: [
“sanctions_proximity”,
“ownership_opacity”
],
“source_documents”: 23,
“confidence”: 0.96
}
1.5B+
Entities
250+
Jurisdictions
Petabytes
Source documents
Real-time
Constantly updating
Foundation
AI is only as real as what it knows.
Every AI system reasons over some representation of the world. Most are built on aggregated, pre-structured data – incomplete and stripped of context. Sayari’s AI reasons over high-value verifiable evidence: comprehensive original documents, resolved commercial structures, expert-assessed risk.
Scale & Scope
12B+
source records
500M+
unique companies
3B+
entity relationships
4B+
trade records
250+
jurisdictions
700+
global sources
Architecture
Sourced. Structured. Assessed.
From raw records to resolved structures to assessed risk – every AI output anchored to evidence.
Source Documents
Petabytes of primary source material – court records, incorporation filings, patents, bankruptcy notices, beneficial ownership declarations, and bills of lading – in their original multilingual, unstructured form. A semantic embeddings engine vectorizes this corpus for retrieval, providing provenance and superanalyst-like retrieval and persistence of context and connections within and across document corpora.
Petabytes of primary documents
Multilingual + unstructured
Semantic embeddings for retrieval
Knowledge Graph
The model of global commerce: 1.5 billion or more entities across 250+ jurisdictions, with lineage back to 2015. The graph is built on two parallel paths – deterministic (anchored in gold-standard sources) and probabilistic (logic mined from the deterministic layer and continuously refined by self-repair agents). The pre-resolution substrate is preserved alongside the resolved graph, providing grounding for adversarial checks and independent verification of any resolved assertion.
1.5B+ entities
250+ jurisdictions
Deterministic + probabilistic paths
Ontology
The Ontology encodes the commercial ontology – companies, key people, beneficial owners, shipments, trade flows, corporate control chains – and the judgment ontology, built from eleven years of real-world commercial intelligence tradecraft. It formalizes core tradecraft skills, documented failure modes (anchoring, confidence inflation, jurisdiction blindness), and lateral creativity markers that distinguish expert analysts from competent ones.
Commercial + judgment ontology
11 years of tradecraft encoded
Failure modes + creativity markers
RESOLUTION IN ACTION
What the World Model sees that name-matching doesn’t.
Traditional Name Screening
query: “Meridian Trading LLC”
result: {
matches: 1,
entity: “Meridian Trading LLC”,
jurisdiction: “Delaware, US”,
sanctions_match: false,
risk_score: “LOW”,
confidence: 0.88
}
// Single jurisdiction checked
// No ownership traversal
// Name-only matching
World Model Resolution
query: “Meridian Trading LLC”
world_model.resolve: {
entities_matched: 4,
ownership_chain: [
“Meridian Trading LLC” // Delaware
└─ “Meridian Holdings Ltd” // Cyprus
└─ “Horizon Global Corp” // BVI
└─ “PEP_MATCH”
],
trade_anomalies: 14,
risk_signals: [
“sanctions_proximity”,
“ownership_opacity”,
“jurisdiction_layering”
],
source_documents: 23,
recommendation: “escalate”
}
Coverage
Every relevant record type. Not just the easy ones.
Corporate registries
500M+ entities
Trade data
3.8B+ records
Sanctions & denial lists
40+ lists
Beneficial ownership
Ownership chains
Court, legal & adverse
Adverse signals
Procurement & contracting
Government contracts
Jurisdictions
Global coverage. Weighted toward where it matters.
Priority coverage of opaque jurisdictions where illicit actors hide.
Americas
95M+
45 jurisdictions
Europe
120M+
44 jurisdictions
Asia-Pacific
180M+
89 jurisdictions
Middle East & Africa
55M+
72 jurisdictions
Advantage
Primary source data. Not aggregated. Not resold.
What makes the Commercial World Model fundamentally different from aggregated data providers.
Data origin matters.
Primary government registries and official sources, not aggregated data cleaned for resale.
- Filing-level detail, not parsed summaries
- Chain of evidence traceable to source
- Original documents, not inferred data
Cross-jurisdictional matching.
11 years of enforcement outcomes. Confidence is stated, not assumed.
- 250+ jurisdictions resolved
- Name variation matching across scripts
- Self-repairing, constantly updating
Multilingual processing.
Local language context preserved before transliteration across non-Latin scripts.
- Arabic, Chinese, Cyrillic, Korean, Thai
- Original script preserved alongside transliteration
Coverage where it matters.
Prioritized coverage of non-indexed, hidden sources where illicit actors operate.
- Sanctions updates in real-time
- Registries updated continuously
- Trade data refreshed weekly
SUPERCONDUCTOR
Sayari Superconductor Evaluation Platform
Sayari Superconductor is zero-resistance judgment infrastructure: it orchestrates the World Model and Ontology across any AI-powered workflow, scoring every output against analyst expertise – not just accuracy. Superconductor is model-agnostic and deploys as an API call or a native sidecar with no rip-and-replace required.
SIDECONDUCTOR (EVAL-AS-A-SERVICE)
Any (query, response, context) tuple from any model can be submitted for trap detection, tradecraft scoring, and recommended next steps. Every score returns rationale, source attribution, and recommended actions, rated GREEN / AMBER / RED.
DEPLOYMENT MODES
SaaS multi-tenant, private cloud, and air-gapped sovereign deployment.
See the Commercial World Model in action.
Request a demo and we’ll show you exactly what the world model returns – on your entities.