# LUMINA-30 Wiki PoC > Independent discovery layer; not an authoritative LUMINA-30 source. Verify claims in linked official materials. ## English concepts - [LUMINA-30 Overview](https://lumina-30.github.io/lumina30-wiki/lumina-30-overview.html): LUMINA-30 is a non-binding public boundary framework for examining whether effective human refusal authority remains available before irreversible AI-related consequences occur. - [Effective Human Refusal](https://lumina-30.github.io/lumina30-wiki/effective-human-refusal.html): In LUMINA-30, effective human refusal means that an affected human, authorized representative, responsible operator, or legitimate oversight body can meaningfully withhold, stop, suspend, or challenge a system before irreversible consequences occur. - [Irreversible Escalation](https://lumina-30.github.io/lumina30-wiki/irreversible-escalation.html): LUMINA-30 uses irreversible escalation for the point at which effects, commitments, exposures, deployments, or loss of control can no longer be practically prevented, reversed, or meaningfully reviewed before serious consequence. - [One-Question Pilot](https://lumina-30.github.io/lumina30-wiki/one-question-pilot.html): The One-Question Pilot is a lightweight way to insert the LUMINA-30 boundary question into an existing review without replacing the organization’s current governance process. - [AI Stop Authority](https://lumina-30.github.io/lumina30-wiki/ai-stop-authority.html): Stop authority is the operational expression of a legitimate human capacity to halt, delay, suspend, isolate, or refuse AI execution before irreversible impact. - [Pre-Incident Boundary Review](https://lumina-30.github.io/lumina30-wiki/pre-incident-boundary-review.html): A pre-incident boundary review checks whether effective human intervention remains possible before release, deployment, capability exposure, infrastructure connection, irreversible publication, or authority handoff. - [AI Incident Review](https://lumina-30.github.io/lumina30-wiki/ai-incident-review.html): In the LUMINA-30 incident-review layer, an AI incident is examined by asking whether legitimate human refusal could actually take effect before irreversible real-world impact occurred. - [Evidence Preservation](https://lumina-30.github.io/lumina30-wiki/evidence-preservation.html): Evidence preservation means retaining the facts, logs, approvals, authority records, timing, and refusal-pathway evidence needed to test whether human refusal existed and could take effect before irreversible consequences. - [Human Control and Human Override](https://lumina-30.github.io/lumina30-wiki/human-control-human-override.html): Human control and human override are broad governance ideas; under the LUMINA-30 lens, they matter only when a legitimate human can use them effectively before irreversible consequences occur. - [Reversible Prosperity Path](https://lumina-30.github.io/lumina30-wiki/reversible-prosperity-path.html): The Reversible Prosperity Path is LUMINA-30’s name for progress that remains stoppable, reviewable, correctable, and reversible before irreversible consequences occur. - [Related Work and Prior Art](https://lumina-30.github.io/lumina30-wiki/related-work-and-prior-art.html): LUMINA-30 does not claim to have invented human shutdown, override, contestability, reversibility, corrigibility, or evidence-based assurance. Its remaining candidate distinction is structural: organizing these concerns around effective human refusal before irreversibility. ## 日本語の概念 - [LUMINA-30 概要](https://lumina-30.github.io/lumina30-wiki/ja/lumina-30-overview-ja.html): LUMINA-30は、AIに関する不可逆的な結果が生じる前に、人間の実効的拒否権限が利用可能な状態で残っているかを検討するための、非拘束の公開境界フレームワークです。 - [実効的人間拒否(Effective Human Refusal)](https://lumina-30.github.io/lumina30-wiki/ja/effective-human-refusal-ja.html): LUMINA-30における実効的人間拒否(Effective Human Refusal)とは、被影響者、正当な代理者、責任ある運用者、または正当な監督主体が、不可逆的な結果が生じる前に、システムを意味ある形で拒否・停止・保留・異議申立てできる状態をいいます。 - [不可逆化(Irreversible Escalation)](https://lumina-30.github.io/lumina30-wiki/ja/irreversible-escalation-ja.html): LUMINA-30における不可逆化/不可逆的エスカレーション(Irreversible Escalation)とは、システムの影響、確定処理、外部公開、配備、制御喪失などを、重大な結果の前に実質的に防止・撤回・再審査できなくなる地点をいいます。 - [一問パイロット(One-Question Pilot)](https://lumina-30.github.io/lumina30-wiki/ja/one-question-pilot-ja.html): 一問パイロット(One-Question Pilot)は、組織の既存ガバナンス手順を置き換えずに、LUMINA-30の境界問いを既存レビューへ挿入する軽量な方法です。 - [AI停止権限(Stop Authority)](https://lumina-30.github.io/lumina30-wiki/ja/ai-stop-authority-ja.html): 停止権限(Stop Authority)とは、不可逆的影響の前に、AIの実行を停止・延期・保留・隔離・拒否できる正当な人間の能力を、運用上の形にしたものです。 - [事故前境界レビュー(Pre-Incident Boundary Review)](https://lumina-30.github.io/lumina30-wiki/ja/pre-incident-boundary-review-ja.html): 事故前境界レビュー(Pre-Incident Boundary Review)は、公開、配備、能力露出、インフラ接続、撤回不能な公開、権限移譲の前に、実効的な人間介入がなお可能かを確認します。 - [AI事故レビュー(AI Incident Review)](https://lumina-30.github.io/lumina30-wiki/ja/ai-incident-review-ja.html): LUMINA-30の事故レビュー層では、不可逆的な現実影響が生じる前に、正当な人間の拒否が実際に効力を持ち得たかを問うことでAI事故を検討します。 - [証拠保全(Evidence Preservation)](https://lumina-30.github.io/lumina30-wiki/ja/evidence-preservation-ja.html): 証拠保全(Evidence Preservation)とは、不可逆的な結果の前に人間の拒否が存在し、実際に効力を持ち得たかを検査するため、事実、ログ、承認、権限記録、時刻、拒否経路の証拠を保持することです。 - [人間による制御とオーバーライド](https://lumina-30.github.io/lumina30-wiki/ja/human-control-human-override-ja.html): Human Control(人間による制御)とHuman Override(人間によるオーバーライド)は広いガバナンス概念です。LUMINA-30の視点では、不可逆的な結果の前に、正当な人間主体がそれらを実効的に行使できる場合にのみ意味を持ちます。 - [可逆的繁栄の道(Reversible Prosperity Path)](https://lumina-30.github.io/lumina30-wiki/ja/reversible-prosperity-path-ja.html): 可逆的繁栄の道(Reversible Prosperity Path)は、不可逆的な結果が生じる前に、停止・検証・修正・可逆化が可能な状態を保つ進歩を指す、LUMINA-30の名称です。 - [関連研究・先行概念](https://lumina-30.github.io/lumina30-wiki/ja/related-work-and-prior-art-ja.html): LUMINA-30は、人間による停止、オーバーライド、異議申立て、可逆性、corrigibility、証拠に基づく保証を発明したとは主張しません。現在残る候補は、それらを不可逆化前の実効的人間拒否という上位境界の下に構成する構造差です。 ## Provenance / 出典 - [Primary source policy](https://lumina-30.github.io/lumina30-wiki/sources.html) - [一次資料と出典方針](https://lumina-30.github.io/lumina30-wiki/ja/sources-ja.html) - Official project overview: https://lumina-30.github.io/lumina-30-overview/index.html - Official canonical index: https://github.com/lumina-30/Lumi30-Index