Consequential AI dependency
R1 — Preconditions observed
Consequential adoption and dependency are increasingly visible, while substitutability varies by sector.
AI RISK SYSTEM · 2026-09-17
Critical systems & systemic dependence
When does widespread AI adoption become a concentration, dependency or cascading operational risk?
Current assessment
R1 — Preconditions observed. The model recognises growing structural dependency without misclassifying that dependency itself as a realised systemic failure.
The top-level realisation state is the furthest validated state reached by at least one monitored pathway. It does not imply every pathway inside Critical systems & systemic dependence has reached R1.
Exposure: X3 — Consequential deployment. Consequence envelope: C4 — Cross-sector / systemic. Control assurance: A1 — Controls specified / implemented.
Monitored pathways
Shared-provider concentration
R1 — Preconditions observed
Common technology providers and shared infrastructure are recognised channels for correlated disruption.
Correlated AI-specific operational failure
R0 — Hypothesised
The mechanism is credible but verified AI-specific cross-organisation failure evidence remains sparse.
Cascading systemic disruption
R0 — Hypothesised
Systemic cascade is a downstream scenario rather than an observed AI-driven outcome in the baseline.
Where the evidence reaches
Consequential adoption and shared-provider dependence are visible; broad AI-driven systemic cascade is not.
Consequential AI adoption
Growing · robust evidence
AI is increasingly deployed in consequential business and public-sector workflows.
What remains uncertain
Depth of dependence and ability to substitute away from systems vary substantially by sector.
What would move this stage
Sector-level measures of dependency and substitution capacity.
Shared provider / infrastructure dependence
Material concern in some sectors · medium evidence
The Bank of England highlights common technology providers, shared software and critical infrastructure as channels for correlated disruption.
What remains uncertain
Cross-sector concentration and fallback readiness are incompletely measured.
What would move this stage
Auditable dependency maps and resilience tests across critical sectors.
Correlated operational failure
Plausible pathway · limited evidence
Shared dependencies can transmit disruption across multiple organisations at once.
What remains uncertain
How often AI specifically becomes the initiating or amplifying cause is not well established.
What would move this stage
Verified multi-organisation events and comparative baseline evidence.
Fallback & recovery capacity
Uneven · limited evidence
Risk frameworks emphasise resilience, fallback and recovery rather than prevention alone.
What remains uncertain
Operational testing and public evidence are highly heterogeneous.
What would move this stage
Measured restoration, fallback and substitution performance under realistic failure conditions.
Cascading systemic disruption
Not established as general AI-driven outcome · limited evidence
Complex-system theory and sector analysis support the mechanism in principle.
What remains uncertain
Scale, probability and AI-specific causal contribution remain uncertain.
What would move this stage
Verified cross-system consequence evidence.
Evidence limiting the assessment
Many deployments retain human fallback, substitution options or limited decision authority, and AI-specific attribution is often weak.
Claim-level evidence
4 claim-level evidence records currently sit beneath this system. They identify the specific proposition each document is being used to support or limit rather than treating a whole report as one finding.
Key sources
- AI Risk Management Framework — NIST
- Preliminary Report: Independent International Scientific Panel on AI — United Nations
- Financial Stability Report — July 2026 — Bank of England
Compare this system with the full current assessment, inspect the dataset summary, or read the methodology.