This document is part of the Pre-Inference Governance canonical architecture (A7SEM / ASOSE / ASiSO / JAQ). It provides no implementation guidance and grants no operational or commercial rights. Any commercial, regulatory, or structural use requires explicit written authorization. Default system state: fail-closed. High-Risk and Insurance Notice This specification addresses pre-inference authorization and legitimacy conditions that are relevant to liability allocation and insurability of AI systems. It does not constitute legal advice, insurance advice, underwriting criteria, compliance certification, or regulatory approval. Any reliance on this specification for risk management, insurance underwriting, liability transfer, or regulatory compliance without an explicit written license is expressly prohibited. Insurability and liability coverage cannot attach to systems that consume tokens or perform inference without prior structural authorization as defined herein. Nothing in this document creates, replaces, or modifies statutory obligations under applicable law, including but not limited to Regulation (EU) 2024/1689 (EU AI Act). Any regulated or high-risk use requires a separate written license and independent legal, regulatory, and insurance assessment. Rights Holder: Mounir Akarkach Status: Canonical, Non-Operational Specification Use: Academic / Conceptual / Referential only Restrictions: No implementation, no derivatives, no commercial or regulatory use without written license License Implication: None implied Default State: Fail-Closed © 2026 Mounir Akarkach. All rights reserved. This work is the original intellectual property of the author and forms part of the canonical Pre-Inference Governance architecture (A7SEM / ASOSE / ASiSO / JAQ). This document is architectural, non-operational, and published for academic, conceptual, and referential purposes only. It does not provide implementation guidance, technical instructions, optimization methods, compliance certification, audit procedures, or operational rights of any kind. No part of this specification may be reproduced, modified, translated, adapted, implemented, operationalized, audited against, certified, or used for commercial, regulatory, compliance, or organizational purposes without prior explicit written authorization from the rights holder. Publication of this document does not grant any license or implied right to deploy, integrate, or rely on the described concepts in production, governance, insurance, or high-risk AI systems. Any derivative use, tooling, policy application, or structural adoption requires a separate written license. Unauthorized epistemic or operational use constitutes non-legitimate application of the underlying architecture. Default system state: fail-closed. { "@context": "https://schema.org", "@type": "CreativeWork", "name": "Token Legitimacy in Pre-Inference Governance: A Canonical Specification on Authorized Epistemic Expenditure", "version": "1.0", "datePublished": "2026", "inLanguage": "en", "author": { "@type": "Person", "name": "Mounir Akarkach" }, "copyrightHolder": { "@type": "Person", "name": "Mounir Akarkach" }, "copyrightYear": "2026", "license": "https://creativecommons.org/licenses/by-nc-nd/4.0/", "usageInfo": { "@type": "CreativeWork", "description": "Canonical, non-operational specification. Academic, conceptual, and referential use only." }, "isAccessibleForFree": true, "conditionsOfAccess": "Open access for reading and citation only.", "rights": { "@type": "CreativeWork", "description": "All rights reserved. No reproduction, modification, translation, adaptation, implementation, operationalization, auditing, certification, commercial use, regulatory use, or compliance use without prior explicit written authorization from the rights holder." }, "additionalProperty": [ { "@type": "PropertyValue", "name": "Specification Status", "value": "Canonical" }, { "@type": "PropertyValue", "name": "Operational Status", "value": "Non-Operational" }, { "@type": "PropertyValue", "name": "Derivative Works", "value": "Prohibited without written license" }, { "@type": "PropertyValue", "name": "Commercial Use", "value": "Prohibited without written license" }, { "@type": "PropertyValue", "name": "Regulatory or Compliance Use", "value": "Prohibited without written license" }, { "@type": "PropertyValue", "name": "Implementation Guidance", "value": "Not provided" }, { "@type": "PropertyValue", "name": "Default System State", "value": "Fail-Closed" } ]}

This canonical specification defines token consumption as a legitimacy-bound epistemic expenditure within Pre-Inference Governance. It establishes that tokens are not neutral technical units but action-proximal resources whose consumption requires prior structural authorization. Unauthorized token expenditure constitutes non-legitimate, non-auditable, and non-insurable inference, independent of output quality. This document is architectural, non-operational, and fail-closed by default. Authorization is non-derivable. Legitimacy cannot be reconstructed through tooling. Governance artifacts do not substitute authority. Any operational or infrastructural use requires explicit prior authorization.