Knowledge Graph & Citations

Two complementary endpoints: a knowledge graph that maps relationships between all entities in the taxonomy, and a citation index with pre-formatted citation strings.

Knowledge Graph

Endpoint: /api/graph.json (opens in new tab)

Format: JSON (UTF-8)

Citation Index

Endpoint: /api/citations.json (opens in new tab)

Format: JSON (UTF-8)

Authentication: None required

CORS: Access-Control-Allow-Origin: *

Cache: public, max-age=3600 (1 hour)

Knowledge Graph

The graph endpoint provides all nodes and typed edges in the TopAIThreats taxonomy. It enables relationship-aware queries across domains, patterns, incidents, and glossary terms.

Node Types

Type	ID Format	Key Fields
`domain`	`domain:DOM-INF`	domain_code, title, definition, url, framework_mapping
`sub_category`	`sub:DOM-INF-slug`	pattern_code, title, severity, likelihood, url
`incident`	`incident:INC-24-0001`	title, status, severity, date_occurred, regions, sectors
`glossary_term`	`glossary:term-slug`	term, definition, url, same_as

Every node includes a degree field indicating the total number of edges (inbound + outbound) connected to it.

Edge Types

Relationship	Direction	Meaning
`has_pattern`	Domain → Pattern	Domain contains this threat pattern
`belongs_to`	Pattern → Domain	Inverse of has_pattern
`primary_pattern`	Incident → Pattern	Primary threat classification
`secondary_pattern`	Incident → Pattern	Secondary threat classification
`references_term`	Incident\|Pattern → Glossary	References a glossary term
`explains`	Glossary → Pattern	Term explains this pattern
`related_pattern`	Pattern → Pattern	Cross-domain relationship
`related_term`	Glossary → Glossary	Semantic relationship

Meta & Diagnostics

The _meta object includes aggregate statistics (stats) and a warnings array flagging coverage gaps (e.g. patterns with no documented incidents, glossary terms with no cross-references).

Example (Graph)

{
  "_meta": {
    "name": "TopAIThreats.com Knowledge Graph",
    "graph_version": "2.0",
    "schema_version": "2.0",
    "generated": "2026-02-26T00:00:00.000Z",
    "stats": {
      "domains": 8,
      "sub_categories": 42,
      "incidents": 56,
      "glossary_terms": 140,
      "total_nodes": 246,
      "total_edges": 512
    },
    "edge_types": [ ... ],
    "warnings": [
      "Coverage gap: Pattern X has no documented incidents"
    ]
  },
  "nodes": [
    {
      "id": "domain:DOM-INF",
      "type": "domain",
      "domain_code": "DOM-INF",
      "title": "Information Integrity",
      "degree": 12,
      ...
    }
  ],
  "edges": [
    {
      "source": "domain:DOM-INF",
      "target": "sub:DOM-INF-deepfake-identity-hijacking",
      "relationship": "has_pattern"
    }
  ]
}

Citation Index

The citation endpoint provides pre-formatted citation strings for all citable entities: incidents, domains, and glossary terms. Each entry includes a stable identifier, canonical URL, and a ready-to-use citation string.

Entity Categories

Category	cite_id Format	Fields
`incidents`	`INC-24-0001`	title, url, date_occurred, last_updated, severity, evidence_level, citation
`domains`	`DOM-INF`	title, url, definition, citation
`glossary`	`glossary:term-slug`	title, url, definition, same_as, citation

Example (Citations)

{
  "_meta": {
    "name": "TopAIThreats.com Citation Index",
    "version": "1.0",
    "generated": "2026-02-26T00:00:00.000Z",
    "total_entities": 204,
    "license": "CC BY 4.0"
  },
  "incidents": [
    {
      "cite_id": "INC-24-0001",
      "type": "incident",
      "title": "Hong Kong Deepfake CFO Video Conference Fraud",
      "url": "https://topaithreats.com/incidents/INC-24-0001-hong-kong-deepfake-cfo-fraud/",
      "severity": "critical",
      "citation": "TopAIThreats, \"Hong Kong Deepfake CFO Video Conference Fraud\" (INC-24-0001), https://topaithreats.com/incidents/INC-24-0001-hong-kong-deepfake-cfo-fraud/, last updated 2026-02-26."
    }
  ],
  "domains": [ ... ],
  "glossary": [ ... ]
}

Use Cases

Graph visualisation — Build interactive network diagrams of the AI threat landscape using nodes and edges
Relationship queries — Traverse edges to find which incidents relate to specific patterns, or which glossary terms explain particular threats
Coverage analysis — Use the warnings array and degree counts to identify under-documented areas
Automated citation — Generate properly formatted references to TopAIThreats content in reports, papers, and briefings
LLM grounding — Provide citation strings as context so language models can reference specific incidents with verifiable URLs

View graph endpoint (opens in new tab) · View citations endpoint (opens in new tab) · All endpoints