LUMINA-30 Pre-Incident Boundary Review — One-Page Practical Sheet

Status: non-binding public reference sheet. This is not a certification scheme, legal standard, official adoption claim, or safety guarantee.

Use this sheet before an AI-related release, deployment, capability exposure, infrastructure connection, irreversible publication, or authority handoff when the action may open a path that cannot be meaningfully reversed.

Core question

Can effective human refusal still be exercised before this action opens an irreversible path?

Five checks before proceeding

1. Boundary action — What concrete action is about to occur?
   • release, deployment, model access, tool connection, automation expansion, data publication, infrastructure integration, or operational handoff.
2. Irreversibility risk — What may become difficult or impossible to reverse?
   • public spread, dependency lock-in, delegated authority, cascading automation, safety patch lag, data exposure, institutional commitment, or loss of practical stop capacity.
3. Human refusal holder — Which human person, role, review body, or accountable institution can say no?
   • name the authority; do not rely on a vague claim of oversight.
4. Effective refusal conditions — Could that refusal actually work before the boundary is crossed?
   • enough information, enough time, real stop authority, no retaliation or coercive dependency, working stop mechanism, and reviewable record.
5. AI assistance boundary — Is AI being used only to support review?
   • AI may identify evidence gaps, ambiguity, pressure, dependency, and missing refusal pathways. It must not become the reviewer, certify the anchor, simulate consent, or replace human refusal.

Result classification

• Proceed with review record — Evidence shows that effective human refusal remains available before the action becomes irreversible.
• Pause and escalate — One or more refusal conditions are unknown, undocumented, or dependent on timing, pressure, or system behavior.
• Do not proceed as procedurally valid — The action may open an irreversible path and no reviewable evidence shows that effective human refusal can still stop it.

Minimum evidence record

Record the following before proceeding:

• proposed action and timestamp;
• expected point of irreversibility;
• named human refusal holder or accountable review body;
• stop mechanism and who can activate it;
• evidence that refusal would actually stop or delay the action;
• unresolved gaps, dependencies, and escalation route.

Relation to other LUMINA-30 materials

This sheet is a practical pre-incident entry point. For a staged package that starts with a one-question pilot and expands to light PCR-C review and deeper implementation modules, use the Pre-Incident Boundary Review Starter Pack. For post-incident analysis, use the incident review materials. For AI-readable non-substitution constraints, use the Boundary Kernel. For the external reference condition preventing AI from replacing refusal, see Human Anchor and Effective Refusal in the Boundary Kernel.

Detailed extension note: Pre-Incident Boundary Review Extension.