Layered Governance Architecture

A three-layer framework integrating model-level, agent-level, and ecosystem-level governance for real-world agentic AI systems with adaptive control.

0.554
Violation Detection Rate
0.0
Detection Latency
1.0
Attribution Accuracy
0.834
Utility Preservation
0.31
Constant Overhead

Framework Comparison

Violation Detection by Framework

Utility vs Overhead Trade-off

Scaling & Adaptation

Scaling: Agents vs Detection Rate

Adaptive Response to Risk Phases

Detailed Results

Framework Comparison

FrameworkViol. RateLatencyAttributionOverheadUtility
None0.0983.3301.00.001.000
Model Only0.4950.01.00.120.852
Agent Only0.5390.01.00.150.838
Ecosystem Only0.2430.01.00.180.927
Layered (LGA)0.5540.01.00.310.834

Scaling Behavior (Layered)

AgentsViol. RateOverheadUtility
20.2050.310.939
40.3350.310.900
80.3540.310.894
160.3570.310.893
320.3700.310.889

Adaptation Under Risk Changes

PhaseViol. RateUtilityMean Risk
Normal0.1680.9500.229
Spike0.8300.7510.526
Recovery0.1940.9420.228