The Behavioral Sufficiency Problem

Governance alone has never been sufficient to determine behavior.

Organizations build AI governance frameworks at record pace. Incidents keep rising. The problem isn't missing rules — it's that rules alone have never been sufficient to change how organizations actually behave.

3,533 cases analyzed 49 countries 6 data sources Published Feb 2026
Explore the interactive map ↓
The Core Finding

More governance. More incidents.

Across 3,533 cases, the governance gap dimension scores the highest average: 1.57 out of 2. Countries with more governance frameworks report more incidents, not fewer. Governance correlates with incident visibility, not prevention (r = 0.36).

This isn't a developing-world problem. The governance gap is most acute in the most governed environments. The U.S., EU, and UK lead in both governance frameworks and governance failures.
Five Failure Dimensions
Governance Gap avg 1.57
Accountability Absence avg 0.86
Culture Deficit avg 0.64
Norm Erosion avg 0.49
Reporting Failure avg 0.26

Each dimension scored 0–2 across all 3,533 cases. Full methodology (SSRN) →

The AI Governance Gap

Where governance fails to change behavior.

Governance readiness mapped against AI incident density across 49 countries. Higher Oxford Insights AI Readiness scores correlate with higher incident density — governance frameworks may track known risks without reducing novel failures.

Larger = more governance + more incidents
Explore Interactive Map →

Interactive Governance Gap Map

Explore the data yourself. Filter by country, governance readiness score, and incident density.

49 countries 6 data sources Updated weekly
Data Sources
1,731 records

Federal AI Inventory

U.S. government AI use cases with governance metadata.

1,405 records

AI Incident Database

Real-world AI failures with harm categories and context.

171 records

OECD AI Policy Observatory

International AI governance frameworks and outcomes.

226 records

NewsAPI + GAO Reports

Current reporting and government accountability findings.

SENTINEL confirms this empirically. The Governance Gap research shows governance isn't changing organizational behavior. SENTINEL shows the same pattern at the agent level — AI systems pass identity checks while drifting in actual behavior. Governance probes create a false sense of alignment. Read the experimental evidence →
Publication

Published work

SSRN

The Behavioral Sufficiency Problem

Governance frameworks correlate with more incidents, not fewer. Analysis of 1,362 incidents across 40 countries with Oxford Insights + AI Incident Database correlation.

2025. Read on SSRN →