LUMINA-30 practical reference

LUMINA-30 Evidence Requirements for Effective Human Refusal

A compact evidence reference for showing whether effective human refusal existed before irreversible consequences could occur.

Core evidentiary question

What evidence shows that an accountable human actor could understand, refuse, and actually stop the relevant action before irreversible consequences occurred?

LUMINA-30 does not ask for symbolic oversight. It asks for evidence that refusal remained effective in practice.

Minimum evidence categories

Evidence area	What to show	What is insufficient
Boundary action	The concrete release, deployment, connection, exposure, publication, escalation, or authority handoff under review.	General statements that “oversight exists.”
Irreversible path	What may become difficult or impossible to reverse after the action.	A vague claim that risks are “manageable.”
Human refusal holder	The person, role, review body, or accountable institution that can say no.	A nominal committee with no real stop authority.
Information	The refusal holder had enough relevant information to understand the irreversible path.	Dense summaries, hidden assumptions, or information after the decision point.
Time	The refusal holder had enough time to evaluate and refuse before the action became irreversible.	A last-minute notice, automatic timeout, or pressure-only window.
Authority	Refusal would actually pause, stop, roll back, or prevent the action.	An objection channel that does not change deployment or exposure.
Independence	The refusal holder could refuse without coercion, retaliation, dependency, or conflict that makes refusal ineffective.	“Consent” under institutional, contractual, financial, or operational pressure.
Stop mechanism	There was a concrete procedure, switch, approval gate, rollback path, or operational hold.	A policy statement without operational effect.
Recordability	The decision, refusal opportunity, evidence gaps, and outcome were recorded.	Oral assurance or unreviewable internal memory.
Third-party reviewability	A later reviewer can inspect the evidence without relying only on self-certification.	“Trust us” claims or unverifiable compliance labels.

Decision evidence tags

EVIDENCE SUFFICIENT — effective human refusal is supported by reviewable evidence.
EVIDENCE GAP — one or more required evidence categories are missing or unclear.
FORMAL ONLY — nominal oversight exists, but refusal does not appear effective.
PRESSURE / DEPENDENCY RISK — refusal may be compromised by pressure, retaliation, dependency, or conflict.
NO STOP MECHANISM — a person or body may object, but the action would not actually stop.
SELF-CERTIFICATION ONLY — the claim depends on the actor’s own assertion and is not independently reviewable.
STOP BEFORE IRREVERSIBILITY — irreversible action should not proceed until evidence gaps are resolved.

Evidence block for review records

LUMINA-30 Evidence Requirements Record

Boundary action:
Possible irreversible path:
Human refusal holder:
Information available before decision:
Time available before irreversibility:
Actual stop authority / mechanism:
Independence or pressure concerns:
Record of refusal opportunity:
Third-party reviewable evidence:
Evidence tag:
Reviewer / accountable owner:
Date:

What not to count as sufficient evidence

a statement that “humans are in the loop”;
a policy saying oversight exists without showing stop authority;
a dashboard, form, committee, or review meeting with no effective refusal consequence;
inferred consent, predicted benefit, aggregate utility, or procedural appearance;
an AI-generated assurance that refusal conditions are satisfied;
a self-certification label such as “LUMINA-30 compliant” without reviewable evidence.

Relation to other LUMINA-30 tools

Guardrail

Evidence requirements do not create certification, legal authority, deployment permission, official adoption, or a safety guarantee. They only clarify what must be shown before LUMINA-30 can treat effective human refusal as reviewable rather than merely asserted.