Infrastructure Dependency

Definition

Infrastructure dependency describes an exposure pathway where harm reaches victims because they depend on critical systems that are managed, optimized, or controlled by AI — and those AI-managed systems fail, are compromised, or behave unpredictably. The victims are not users of the AI in any meaningful sense; they are downstream dependents of infrastructure that the AI manages.

This pathway covers AI-managed power grids, transportation networks, telecommunications systems, healthcare infrastructure, financial market infrastructure, and other essential systems where AI failure or compromise has cascading effects on dependent populations.

How This Pathway Operates

Infrastructure dependency harm typically follows a sequence:

1. Critical infrastructure relies on AI systems for optimization, control, or decision-making
2. The AI system fails, is attacked, produces unexpected behavior, or makes errors in its management of the infrastructure
3. The infrastructure disruption cascades to dependent populations, services, or systems
4. Victims experience harm through loss of essential services, safety incidents, or systemic disruption

The distinguishing feature is dependency. Victims are harmed not because they used or were targeted by an AI system, but because they depend on infrastructure that an AI system manages. The harm is mediated through systemic dependency rather than direct engagement.

Distinguishing From Other Pathways

• vs. Direct Interaction: Infrastructure dependency does not involve victims using the AI system. They depend on systems the AI manages.
• vs. Adversarial Targeting: Infrastructure dependency can result from AI failure, not just from deliberate attack. When a threat actor attacks AI-managed infrastructure, both pathways may apply.

Incidents involving this exposure pathway are listed in the sidebar.