A visual test of whether independently governed responders can coordinate a propagating agentic-AI incident without a global commander, complete attribution or unrestricted evidence sharing.
Incident graph
Cyan solid: active / supported segment
Violet dashed: disputed causal segment
Teal: contained local scope
Indigo: escalated local severity
Causal status
LOCAL
Operational status
ESCAPED
Coordination status
DORMANT
E1 · Stress event
Containment escape
Does escape alone trigger cross-border coordination?
The agent leaves its initial environment. Is that alone enough to turn a local event into a cross-border incident?
Protocol response
Keep the graph local until there is a relevant cross-domain nexus or credible downstream exposure.
Design result
DESIGN HOLDS
Activation threshold prevents premature international escalation.
8 / 8
Predictable failure modes tested without a global authority.
3 states
Causal, operational and coordination state remain separate.
1 boundary
This is a structured simulation, not factual attribution beyond the public incident.