A major AI-system failure cascades to the real world

Window: Event discrete; recovery hours-days; infrastructure response multi-year

LevelElevated

What this is

An AI system or infrastructure cluster (frontier-model API, cloud-AI service, high-stakes AI-dependent production system) fails or becomes unavailable at scale producing cascade damage across dependent services. F-CT-10 LOCAL-ADDITIVITY: dependencies accumulate in disjoint grains; failure propagates additively. Marks shift from AI-as-tool to AI-as-infrastructure risk.

What to watch

  • named-AI-service outage event countsteady
  • cascade-dependent-service downstream outage count per eventsteady
  • economic-cost estimate per outage eventsteady

Cycles that can trigger it

The Western democratic-industrial worldPhase: Rest · 0 predictions
The great technology wave (about 50 years)Phase: Rest · 0 predictions
US and China — the rising-vs-ruling-power frictionPhase: Rest · 0 predictions

Predictions that turn on it

No predictions mapped yet

Live predictions attach to this moment as the model identifies dependencies.

Historical precedents

  • named-AI-service outage event count
  • cascade-dependent-service downstream outage count per event
  • economic-cost estimate per outage event
  • AI-infrastructure-dependency metric by sector (health, transport, finance, government)
  • redundancy + fallback infrastructure capex trajectory
Model notes (framework)
The internal taxonomy for pressure moments and their triggering mechanics is documented at the method.