Changelog
A record of significant updates to TopAIThreats.com. For individual incident updates, see each incident's Update Log. For methodology changes, see the Methodology page.
February 2026 — API & Data Access Hub
- Created API & Data Access hub with 4 documentation pages covering all 6 machine-readable endpoints
- Documented Taxonomy API (
/api/threats.json), Incidents API (/api/incidents.json), Knowledge Graph & Citations (/api/graph.json,/api/citations.json), and Feeds & LLM Integration (/rss.xml,/llms.txt) - Added API & Data to header navigation
- Added cross-links between Resources and Governance Frameworks pages
- Added ISO/IEC 42001 to Resources page
- Added "Other Frameworks" cross-navigation to framework detail pages
- Total built pages: 388
February 2026 — Cross-Reference Population
- Populated
causal_factors,assets_involved, andattack_lifecycleon all 56 incidents - Added
risk_assessment(impact scope and reversibility) to all critical and high severity incidents - Populated
causal_factors,assets_involved, andtypical_lifecycle_stageson all 42 threat patterns - All cross-references use validated slugs from the taxonomy's causal factors (15), assets (12), and lifecycle stages (6)
February 2026 — Taxonomy Expansion
- Added taxonomy dimensions: Causal Factors (15), Harm Types (7), Assets & Technologies (12), Governance Frameworks (3), and Risk Levels
- Created index and detail pages for all new dimensions
- Extended incident and pattern schemas with optional cross-reference fields
February 2026 — ID Migration
- Modernised incident ID format from
AIT-YYYY-NNNNtoINC-YY-NNNN - Introduced domain codes (
DOM-INF,DOM-SEC, etc.) and pattern codes (PAT-INF-001, etc.) - Renamed all 56 incident files, updated all frontmatter, schemas, validation, APIs, and cross-references
- Pattern codes are permanent and displayed on pattern detail, pattern index, domain detail, taxonomy, and incident sidebar pages
February 2026 — Ontology Restructure
- Promoted sub-categories to Threat Patterns at top-level URLs (
/patterns/[slug]/) - Renamed Prevention to Affected Groups and expanded from 7 to 9 groups
- Created Sectors index with 18 sector detail pages
- Rewrote Entity registry: replaced heuristic keyword extraction (~190 noisy pages) with curated organisation-only extraction (~40 entities)
- Added
victim_organizationsandthreat_actorsfields to incident schema - Incident sidebar now links to entity pages, sector pages, and affected group pages
February 2026 — New Incidents
- INC-23-0016: Bing Chat (Sydney) System Prompt Leak
- INC-25-0004: EchoLeak — M365 Copilot Zero-Click Prompt Injection (CVE-2025-32711)
- INC-25-0005: ChatGPT Windows Keys Jailbreak
- INC-25-0006: ChatGPT Shared Links Data Exposure
- INC-25-0007: GitHub Copilot RCE via Prompt Injection (CVE-2025-53773)
- INC-25-0008: Cursor IDE CurXecute & MCPoison MCP Vulnerabilities (CVE-2025-54135/54136)
- Total incidents: 56 (23 open, 33 resolved)
January 2026 — Content and Structure Update
- Expanded About, Methodology, Privacy, Terms, Contact, and Contributing pages with full content
- Upgraded threat pattern pages to 3-column layout with metadata sidebar and table of contents
- Merged framework and incident methodology into a single comprehensive Methodology page
- Added cross-domain related patterns to pattern sidebars
- Improved mobile responsiveness across all page templates
2025-01-15 — Initial Launch
- Published 8 threat domains with full definitions and framework mappings
- Published 42 threat patterns across all domains
- Documented 50 initial incidents with primary-source verification
- Launched taxonomy page with full domain and threat pattern structure
- Published incident ledger with filtering and stable incident identifiers
- Schema.org JSON-LD markup on all pages
- Machine-readable API endpoints: /api/threats.json and /api/incidents.json
- LLM reference guide at /llms.txt
- Sitemap generation for search engine indexing
Changelog Scope
The changelog records:
- Taxonomy changes
- Methodology updates
- Structural or classification revisions
- Major content corrections
Routine editorial edits may not be listed.
Versioning Philosophy
Top AI Threats is maintained as a living reference. Changes prioritise transparency and traceability over static completeness.