Design validation This is a detector demo on synthetic fixtures — not a conformity assessment, not an ISO 42001 compliance product, and not certification. EVE surfaces a potential degradation of decision-chain coherence. EVE does not classify incidents and makes no compliance determination. All material is synthetic; no customer data is used; nothing is sealed.
01 · Framing

The trigger is an accountability gap

The trigger is not automation level. The trigger is an accountability gap. A fast, highly automated chain with clear declared ownership may not need a checkpoint; a slow chain with diffuse ownership does. A checkpoint is required when declared accountability for the next decision cannot be confirmed.

Reviewer-informed design note. See ADR-013. Delay, automation level and complexity are contributing indicators only — never sufficient trigger conditions on their own.

02 · Detector demo

Run the detector on a synthetic case

checkpoint_required:
Case
Trigger basis
Accountable owner
Declared authority
Authority validity
Last confirmed human review
Changed facts since review
Required human confirmation before the chain proceeds
    
    
    
  
    

  
  
    03 · Decision chain
    
  
    Where the checkpoint sits
    
  
    
    
    1 · Agent
    Agent recommendation
    
    
    2 · System
    System enrichment
    
    
    3 · Human
    Human review
    
    
    4 · Change
    Scope / risk change
    
    
    5 · Checkpoint
    Next decision requires accountability confirmation
    
  
    

  
  
    04 · Boundary
    
  
    What this is, and is not
    
  
    
    
    Responsibility boundary
    
    
    EVE does not approve, reject, classify an incident, or determine compliance. EVE surfaces a potential degradation of decision-chain coherence. A named human owner must confirm authority, responsibility and next step before the chain proceeds.