Workbench / Issues

Model Output Triggers Unapproved Action

AI output causes, recommends, or triggers an action that has not passed the required approval, permission, or authority check.

What This Looks Like

The AI produces an output that causes an action to happen before the required approval or permission check is complete. A generated field, recommendation, classification, confidence score, tool instruction, or routing result may trigger a workflow action that should have required human or system approval first.

Why It Matters

AI output can become operational input. If an output triggers action without approval, the model effectively gains authority it was not supposed to have. This can create unauthorized workflow changes, escalations, notifications, tool calls, or downstream decisions.

Structural Signal

The output crosses from recommendation or classification into action without passing through the required authority boundary. The issue is not only that the action may be wrong; it is that the approval structure was bypassed.

Common Triggers

When to Use This Issue

Use this Issue when AI output triggers or enables an action that should have required approval, permission, or declared authority first.

When Not to Use This Issue

Do not use this Issue when the AI merely suggests an action and no workflow step is triggered. Do not use it when approval exists and is properly applied. This Issue applies when model output crosses into unapproved action.

Category

Permissions & Approvals

Primary Pattern

PAT-0420 — Undeclared Side Effect

Declared Patterns

Derived Primary Lenses

Derived Secondary Lenses

Search Intents

Ontology Metadata

Code
ISS-0050
Version
ISS-0050@0.1.0
Ontology release
0.1.0
Updated
2026-05-10T00:00:00Z

History

No public history entries recorded.

View full ontology changelog