Adversarial Targeting

Definition

Adversarial targeting describes an exposure pathway where a threat actor deliberately uses AI as a weapon to harm specific victims. The AI system is not simply malfunctioning or producing unintended consequences — it is being actively directed against identified targets.

This pathway encompasses deepfake fraud (using synthetic media to impersonate executives), AI-powered spear phishing (generating personalized social engineering attacks), AI-enhanced malware (using machine learning to evade detection), and other cases where AI amplifies the effectiveness of intentional attacks.

How This Pathway Operates

Adversarial targeting typically follows a sequence:

1. A threat actor identifies specific targets (individuals, organizations, or systems)
2. The threat actor uses AI capabilities to craft, enhance, or scale the attack against those targets
3. The AI-enabled attack reaches the victims through social engineering, technical exploitation, or information manipulation
4. Victims experience harm as a direct result of the targeted attack

The distinguishing feature is intentionality. The threat actor chose to use AI to attack specific victims. This contrasts with direct interaction (where the victim uses an AI system that happens to cause harm) and algorithmic decision impact (where an institutional AI makes harmful decisions about people).

Distinguishing From Other Pathways

• vs. Direct Interaction: Adversarial targeting requires a threat actor who weaponizes AI. Direct interaction involves AI systems operating as designed (even if poorly).
• vs. Infrastructure Dependency: Adversarial targeting is about using AI as a weapon. Infrastructure dependency is about failing AI-managed systems that people depend on.

Incidents involving this exposure pathway are listed in the sidebar.