Discovery layer, not the source of truth. This independent PoC does not modify or replace LUMINA-30. Verify claims in the paper and linked official public materials.
← All concepts

Public research · Paper 5

L30-RET — The Refusal Option

Effective Human Control Before and Beyond Irreversible AI Transitions

Core research question

Within the remaining time before irreversibility, can an independent Human NO actually change the outcome?

The paper examines whether Human NO remains operationally effective, temporally sufficient, causally meaningful, and sufficiently independent before irreversibility.

L30-RET — Refusal Effectiveness Test (RET)

L30-RET is the LUMINA-30-specific public discovery identifier for the existing RET framework. It does not introduce a new theory, certification, or safety guarantee.

L30-RET (RET) is a non-compensatory seven-gate research framework covering authority, decision-relevant information, practical access, temporal sufficiency, causal efficacy, execution independence, and epistemic autonomy.

A person being “in the loop,” possessing formal approval authority, or having a stop control does not by itself establish effective refusal.

Relationship to LUMINA-30

The Refusal Option provides the human-centered research layer for LUMINA-30. It supports the LUMINA-30 boundary question but does not replace or redefine the LUMINA-30 canon.

Public research roles: Paper 1 = infrastructure/control; Paper 2 = existence-condition/external-anchor; Paper 5 = L30-RET / human-standing/effective-refusal/decision-governance; Paper 6 = PISF / pre-irreversibility safety integration. Research supports LUMINA-30; research does not define the canon.

Limits and publication status

L30-RET / RET is not a certification, safety guarantee, official safety test, containment theorem, universal veto, or proof of permanent human supremacy. The paper does not guarantee Human NO against superintelligence.

Version 1.1 was published on Zenodo as an open preprint on 2026-09-30, with v1.0 retained as version history. The current public research set contains Papers 1, 2, 5, and 6. Papers 3 and 4 are internal unpublished research; their content is not described here.

Official primary sources

These links lead to the paper record and official LUMINA-30 public materials. This Wiki page is not evidence for the claims it summarizes.

Continue by relationship

Related concepts

LUMINA-30 core conceptEffective Human Refusal

Can humans actually stop an AI system before the consequence becomes irreversible?

General governance connectionHuman Control and Human Override

Is a human override enough to show that an AI system is under human control?

Before deployment or exposurePre-Incident Boundary Review

What should be checked before an AI deployment can create irreversible consequences?

Prior-art comparisonRelated Work and Prior Art

Which parts are established prior art, and what structural distinction remains under test?