Finite-Capacity AI Constitution — Companion Expansion v1.1

This record is the Companion Expansion v1.1 of the Finite-Capacity AI Constitution. The ZIP file contains the complete companion bundle, including the README and the four individual instruments. The combined PDF is uploaded separately for immediate preview and continuous reading. The companion bundle supports the Finite-Capacity AI Constitution: Canonical Document Package v1.0.1 (April 2026). The canonical package states the constitutional core; this companion expansion makes that core more institutionally readable, inspectable, and usable without reopening or revising it. This expansion was created because a constitutional proposal is not sufficient if readers cannot determine where to begin, how failures are disclosed, who has authority to review or amend claims, and what evidence surface would make those claims inspectable. The v1.1 companion instruments therefore provide navigation, failure handling, authority discipline, and implementation-facing templates around the stable constitutional core. The Finite-Capacity AI Constitution is a structural governance proposal for AI systems operating under bounded resources, incomplete review, uncertainty, and pressure. It does not ask institutions to trust AI systems more. It asks them to require admissibility before execution, explicit boundary acts, reviewable witness, declared authority, and failure disclosure before trust. The Companion Expansion v1.1 contains four instruments: 1. Reader’s Guide / Start Here A navigation guide explaining where different readers should begin and what they do not need to read. 2. Failure and Disclosure Protocol A protocol for making constitutional breakdown reviewable rather than opaque. Its central principle is: failure must not disappear. 3. Authority, Amendment, and Human Review Protocol A protocol defining declared authority, non-implied approval, amendment lineage, meaningful human review, escalation, and redress. Its central principles include: human silence is not approval; human review is meaningful only where non-approval remains possible. 4. Implementation / Evidence Template Pack Illustrative templates for minimal reviewable evidence packages, including examples of manifest.json, actions.jsonl, frames.jsonl, certificate.json, failure_report.json, authority_declaration.json, amendment_record.json, checksums.sha256, and a minimal validator skeleton. This companion expansion is not a legal code, a complete ethics framework, or a compliance guarantee. It does not prove legal compliance, ethical legitimacy, universal safety, fairness, deployment readiness, or institutional adequacy in all contexts. Its narrower claim is that responsible AI governance remains structurally fragile unless systems can hold, refuse, act, disclose failure, declare authority, and leave evidence for review under finite capacity. Public-facing anchors: Human silence is not approval. Failure must not disappear. Review comes before trust.