AI Threat Domains

The AI Threat Domains framework is a harm-based taxonomy that classifies AI-enabled risks into 8 domains and 48 empirically grounded threat patterns to support regulatory, policy, and operational risk assessment.

All Domains

DOM-AGT

Agentic & Autonomous Threats

Threats caused by AI systems that act independently, persist over time, or coordinate with other systems.

7 threat patterns

DOM-CTL

Human–AI Control Threats

Threats arising from how humans rely on, defer to, or lose control over AI systems.

5 threat patterns

DOM-ECO

Economic & Labor Threats

Threats that distort markets, labor conditions, or the distribution of economic power.

5 threat patterns

DOM-INF

Information Integrity Threats

Threats that undermine the reliability, authenticity, or shared understanding of information.

6 threat patterns

DOM-PRI

Privacy & Surveillance Threats

Threats involving unauthorized inference, tracking, or monitoring of individuals or groups.

5 threat patterns

DOM-SEC

Security & Cyber Threats

AI-enabled attacks that compromise the integrity, confidentiality, or availability of digital systems — through input manipulation, model exploitation, or automated offense.

9 threat patterns

DOM-SOC

Discrimination & Social Harm

Threats that result in unfair treatment, exclusion, or social harm to individuals or groups.

5 threat patterns

DOM-SYS

Systemic & Catastrophic Risks

Threats that emerge from scale, coupling, and accumulation rather than single failures.

6 threat patterns

Framework Overview

How this taxonomy is structured

This taxonomy organises AI-enabled threats by their observable impact, rather than by underlying technical mechanisms.

Threats are grouped into 8 domains, each representing a distinct category of harm
Each domain contains 4–7 threat patterns describing concrete harm mechanisms
Domains are analytically distinct but not strictly exclusive
Cross-domain overlaps are explicitly referenced while maintaining a primary classification

How Domains Interrelate

AI-enabled threats commonly span multiple domains rather than occurring in isolation. Each incident is assigned a primary domain based on its dominant harm, with secondary domain intersections explicitly noted.

Example: A deepfake used for financial fraud is primarily classified under Information Integrity, while intersecting with Security & Cyber and Privacy & Surveillance.

This approach preserves analytical clarity while accurately reflecting how AI-related harms manifest across real-world contexts.

Methodology

Threat patterns are identified through systematic review of academic literature, regulatory filings, incident databases, and primary source reporting. Each pattern undergoes evidence-level assessment before inclusion.

→ View full Methodology

Last reviewed: 2026-03-20 · 8 domains · 48 patterns · View full taxonomy