Feeds & LLM Integration
Two non-JSON endpoints for consuming TopAIThreats content: an RSS feed for syndication and update monitoring, and a plain-text reference document optimised for large language models.
Authentication: None required for either endpoint.
RSS Feed
The RSS feed provides a chronological list of all documented incidents, sorted by most recently updated. It conforms to the RSS 2.0 specification (opens in new tab) and includes Atom self-link for feed discovery.
Channel Metadata
| Field | Value |
|---|---|
title | TopAIThreats.com — Incident Feed |
link | https://topaithreats.com |
language | en-us |
atom:link | Self-referencing link for feed readers |
Item Structure
Each <item> represents one incident and includes:
| Element | Description |
|---|---|
<title> | Incident title |
<link> | Canonical incident URL |
<guid> | Permanent link (isPermaLink="true") |
<pubDate> | Last updated date (RFC 2822) |
<description> | Title, severity, status, date, and primary domain |
<category> | Primary domain slug |
Example (RSS)
<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>TopAIThreats.com — Incident Feed</title>
<link>https://topaithreats.com</link>
<language>en-us</language>
<atom:link href="https://topaithreats.com/rss.xml"
rel="self" type="application/rss+xml" />
<item>
<title>Hong Kong Deepfake CFO Video Conference Fraud</title>
<link>https://topaithreats.com/incidents/INC-24-0001-hong-kong-deepfake-cfo-fraud/</link>
<guid isPermaLink="true">https://topaithreats.com/incidents/INC-24-0001-hong-kong-deepfake-cfo-fraud/</guid>
<pubDate>Wed, 26 Feb 2026 00:00:00 GMT</pubDate>
<description>Hong Kong Deepfake CFO Video Conference Fraud.
Severity: critical. Status: confirmed.
Domain: information-integrity.</description>
<category>information-integrity</category>
</item>
</channel>
</rss>LLM Reference Document
The /llms.txt endpoint returns a structured plain-text document designed for ingestion by large language models. It follows the emerging llms.txt convention (opens in new tab) and provides a comprehensive overview of the entire site.
Content Sections
| Section | Description |
|---|---|
| What This Site Is | Site purpose, audiences, and content principles |
| How This Site Is Structured | Content hierarchy and aggregate counts |
| Domains and Sub-Categories | All 8 domains with their 42 patterns (codes, severity, definitions) |
| Incidents | All 56 incidents with date, severity, status, evidence level, regions, sectors |
| Glossary Terms | All 140 terms with definitions |
| Causal Factors | 15 causal factors with categories |
| Harm Types, Assets, Lifecycle, Frameworks | Additional taxonomy dimensions |
| URL Patterns | All URL patterns for programmatic access |
| Machine-Readable Endpoints | List of all API endpoints |
| How to Cite | Citation format and stable ID scheme |
| Structured Data | Schema.org markup present on pages |
Design Rationale
The llms.txt file is generated at build time from the same content collections that power the website. This ensures consistency between the human-readable site and the machine-readable reference. The plain-text format avoids HTML parsing overhead and is suitable for inclusion in LLM system prompts, retrieval-augmented generation (RAG) pipelines, and automated knowledge bases.
Use Cases
- Feed readers & monitoring — Subscribe to
/rss.xmlin any RSS reader to receive notifications when incidents are added or updated - SIEM integration — Ingest the RSS feed into security information and event management platforms for threat intelligence
- LLM system prompts — Include
/llms.txtcontent as grounding context for AI assistants answering questions about AI threats - RAG pipelines — Use the structured text as a retrieval source for augmented generation about AI safety and risk
- Knowledge base seeding — Import the full taxonomy and incident list into organisational wikis or knowledge management systems