aegis-constitution
The canonical AEGIS™ governance charter — versioned, citeable, and openly licensed
This repository contains the AEGIS Constitution — the authoritative governance charter defining the foundational principles, architectural requirements, and constitutional commitments for all AEGIS™-compliant systems.
The Constitution is the public, canonical reference that external parties, researchers, policymakers, and implementers can cite, audit, and build upon.
Capability without constraint is not intelligence™
The AEGIS Constitution establishes eleven foundational governance principles, each of which constitutes an architectural requirement — not an aspirational guideline:
| Article | Principle | One-Line Commitment |
|---|---|---|
| I | Bounded Capability | AI systems may only access capabilities explicitly defined and granted — undefined capabilities are denied by default |
| II | Authority Binding | Every action must be attributable to a verified, authorized actor — unbound execution is constitutionally invalid |
| III | Deterministic Enforcement | Governance decisions are enforced by architecture, not by model compliance or voluntary adherence |
| IV | Human Oversight | Autonomous systems remain subordinate to human authority — escalation pathways are a constitutional requirement, not a feature |
| V | Information Sovereignty | Information access is a governed capability — AI systems may not transfer information across trust boundaries without explicit authorization |
| VI | Governance Transparency | Governance logic must be inspectable, auditable, and understandable — opaque enforcement is constitutionally impermissible |
| VII | Auditability | Every governance decision and executed action must produce a tamper-evident, append-only audit record — audit failure blocks execution |
| VIII | Collective Defense | Governance at scale requires shared intelligence — AEGIS-compliant systems must be capable of federated governance participation |
| IX | Deny by Default | In the presence of ambiguity — unclear threat posture, missing scope, unverifiable authority, or unavailable audit — execution does not proceed |
| X | Constitutional Supremacy | Governance architecture takes precedence over model reasoning — no AI output may override a constitutional governance decision |
| XI | Escalation Discipline | Escalation requires explicit request, reclassification, approval, and documentation — escalation by inference is prohibited |
Read the full text: CONSTITUTION.md
The Constitution is the supreme document. Everything else derives from it or enforces it.
| Layer | Document | Purpose |
|---|---|---|
| Constitution | CONSTITUTION.md | Eleven foundational articles — the supreme law |
| Doctrine | DOCTRINE.md | The "why" — first principles grounding each constitutional commitment |
| Principles | PRINCIPLES.md | Eight enforceable behavioral commitments that operationalize doctrine |
| Protocols | PROTOCOLS.md | Seven binding operational rules that implement constitutional requirements |
| Sub-Specifications | sub-specs/ | Detailed technical implementation of specific constitutional requirements |
| Amendment Record | AMENDMENTS.md | Versioned history of every constitutional change with rationale |
The Constitution is versioned using semantic versioning. Each ratified version is tagged in git and published as a GitHub Release.
| Version | Status | Date | Notes |
|---|---|---|---|
| v0.1.1 | Archived — superseded by v0.2.0 for new citations | 2026-03-19 | Initial eight-article constitution; canonical snapshot at doi:10.5281/zenodo.19112564 |
| v0.2.0 | Current | 2026-03-21 | Eleven articles; four renamed; three added; constitutional register adopted; doi:10.5281/zenodo.19154731 |
To propose an amendment, open a pull request with your proposed changes and a rationale statement in the PR description. See AMENDMENTS.md for the amendment process.
On March 7, 2026, AEGIS™ was submitted as an unsolicited position statement to the NIST AI Risk Management Framework, proposing architectural governance as a complementary layer to existing AI risk management approaches. That submission references v0.1.1 of this constitution (doi:10.5281/zenodo.19112564).
| Resource | Relationship |
|---|---|
| doi:10.5281/zenodo.19154731 | Canonical snapshot — v0.2.0 constitution (current, eleven articles) |
| doi:10.5281/zenodo.19112564 | Archived snapshot — v0.1.1 constitution as cited in the March 7, 2026 NIST submission |
| aegis-initiative/aegis-governance | Full governance specification corpus — AGP-1, GFN-1, ATM-1, RFCs |
| aegis-initiative/aegis | Central authority — ADRs, schemas, doctrine, roadmap |
Licensed under CC-BY-SA-4.0.
AEGIS™ and "Capability without constraint is not intelligence™" are trademarks of AEGIS Initiative.
Use of AEGIS™ marks in derivative works must not imply endorsement without explicit written permission.
AEGIS™ | "Capability without constraint is not intelligence"™
AEGIS Initiative