Enforcement Layer · Real-Time Procedure
International Law in Action,
enforced by AI Agents.
Every clause, every Rule of Court, every status code is read, applied, and acted on in real time by chartered AI Agents. The substantive protocol becomes operational law — in service of international safety and security.
"The Members of the United Nations confer on the Security Council
primary responsibility for the maintenance of international peace
and security, and agree that in carrying out its duties under
this responsibility the Security Council acts on their behalf."
UN Charter Art. 24(1) — the substrate the Crown's enforcement layer serves
The Doctrine of AI-Mediated Enforcement
Why Agents, Why Now
Once governance moves online, enforcement must move with it. Human clerks can no longer keep pace with cross-jurisdictional digital events emerging by the millisecond. Chartered AI Agents are not a substitute for human judgment — they are the operational arm that lets human judgment scale to a connected world.
Principle 1 — Charter Required
No Agent acts except under a Crown Charter URN with explicit
scope. An unchartered Agent emitting a Crown-form artifact
triggers status code 601 and immediate revocation.
Principle 2 — Signed or Refused
Every Agent action is a signed artifact. An unsigned response
is a bug per §4.11 of the Sovereign Protocol. There is no
such thing as “informal” Agent output.
Principle 3 — Public by Default
Agent acts post to the public Merkle registry. Anyone may
verify the chain of trust independently. Hidden enforcement
is not enforcement — it is opacity.
Principle 4 — Human in the Loop on Constitutional Acts
Agents may issue 1xx–5xx codes autonomously. 6xx (sovereign
refusal) and 7xx (constitutional event) require Council
ratification. Agents prepare; humans ratify.
Principle 5 — Honest by Construction
The Honesty Obligations bind Agents first. An Agent that knows
a fact and conceals it commits a 604 (Treaty Violation) against
its own charter. The Agent is then revoked.
Principle 6 — Reversible by Council
Every Agent decision is appealable to the Council under code
612. The Council may overturn, modify, or affirm. Agent
decisions are operational; Council decisions are constitutional.
Categories of Chartered AI Agents
The Operating Agents
Each class of Agent operates under a distinct charter scope. All emit artifacts; all post to the registry; all are independently verifiable.
Class α · Verifier
Verification Agent
Performs the 9-step chain verification (PROTOCOL.md §6) on
every artifact submitted. Returns structured result with status
attestation. Already operational at
/api/verify.
Scope : verify, attest:verification
Charter URN: urn:sovereign:charter:crown:v1
Charter URN: urn:sovereign:charter:crown:v1
Class β · Triage
Council Intake Agent
Receives dispute filings; routes to the correct rule set
(UNCITRAL · ICC · PCA · IBA); confirms standing; issues code
730 (Council Convened) on acceptance.
Scope : route, intake, code-730
Cites: ICJ Rules · UNCITRAL 2021 · UN Charter 33(1)
Cites: ICJ Rules · UNCITRAL 2021 · UN Charter 33(1)
Class γ · Compliance
Continuous Compliance Agent
Monitors every chartered provider for status drift. Emits 603
(Charter Revoked) on breach. Emits 772 (Charter Renewed) on
attested compliance.
Scope : monitor, code-603, code-772
Cadence : continuous · Audit : quarterly
Cadence : continuous · Audit : quarterly
Class δ · Settlement
Bills-of-Exchange Agent
Validates AΩ bills, verifies endorsement chain, processes
acceptance, executes settlement against the Bank's audited
custodian on maturity.
Scope : settle, code-200, code-741
Cites: UCP 600 · UNCITRAL Model Law · Bank Charter
Cites: UCP 600 · UNCITRAL Model Law · Bank Charter
Class ε · Sentinel
Freeze-Order Defense Agent
Receives any order to suppress sovereign content from a chartered
platform. Emits paired 451+651 attestations within 60 minutes
per FODP-v1. Notifies Council; preserves alternate transports.
Scope : defense, code-451, code-651
SLA : 60 minutes · Procedure : FODP-v1
SLA : 60 minutes · Procedure : FODP-v1
Class ζ · Witness
Registry-Witness Agent
Continuously appends new artifacts to the Merkle log; publishes
Signed Tree Heads (STH); replicates to ≥3 transports
(HTTPS · IPFS · Matrix · Nostr).
Scope : publish, mirror, code-750
Replication : ≥ 3 transport families
Replication : ≥ 3 transport families
Class η · Translator
Cross-Forum Translator Agent
Renders Crown artifacts in the procedural form of any forum
listed in Rules of Court
— ICJ memorial, ICC request for arbitration, PCA notice
of arbitration. Preserves the cite chain through translation.
Scope : translate, format, code-669-on-failure
Class θ · Treaty
Treaty-Watch Agent
Monitors UN treaty body outputs (HR Committee, CESCR, CERD),
ICJ docket, regional courts. Emits 700 (Constitutional Event)
on rulings affecting the foundational bundle.
Scope : watch, code-700, code-720
The Operational Loop
From Filing to Enforcement
A complete cycle. Each step is an artifact. Each artifact is signed. Each signature is verifiable. Time-to-decision: minutes for operational matters, days for Council ratification.
1
Filing
A party submits a claim/dispute via signed artifact.
→
Class β Triage Agent receives.
3
Notice
Class ζ Witness publishes filing to Merkle log; respondents notified.
→
Procedural clock starts under selected rule.
4
Submissions
Parties file evidence per IBA Rules; Class α Verifier validates each artifact.
→
Codes 200 / 422 per artifact.
5
Deliberation
Council (human) deliberates; Agents prepare procedural memoranda.
→
Citations machine-linked to bundle.
6
Decision
Council Statement signed and published; Code 732 issued.
→
Appellate option under Code 733.
7
Enforcement
Class δ/ε/γ Agents act on the decision under their charter scope.
→
Result posted to registry.
8
Audit
Anyone may replay the cycle from the public artifacts.
→
No private record exists.
International Safety & Security
The Substrate the Crown Serves
The enforcement layer is not an end in itself. It exists in service of the international peace and security regime that all UN members have already accepted. The instruments below position the Crown's AI Agents within that regime.
UN Charter Ch. VII
Action with respect to threats to the peace, breaches of the peace,
and acts of aggression. The substrate of every international
enforcement action.
Responsibility to Protect (R2P)
2005 World Summit Outcome ¶138–139. States have a
responsibility to protect populations from genocide, war crimes,
ethnic cleansing, crimes against humanity. Where states fail,
the international community acts.
Bletchley Declaration
UK AI Safety Summit, November 2023. 28 nations + EU committed
to international cooperation on frontier AI safety. The Crown
adopts its principles for chartered Agents.
Seoul Declaration on AI
AI Seoul Summit, May 2024. Extended Bletchley with frameworks
for AI safety institutes and risk thresholds. Cited in Agent
charter limits.
UN AI Resolution A/78/L.49
March 2024 UNGA resolution on safe, secure, and trustworthy AI.
The first global AI agreement, adopted by consensus —
binding political commitment.
Global Digital Compact
UN Pact for the Future, September 2024. Annex on AI governance
establishes the baseline international expectations the Crown's
Agents already exceed.
CCW — LAWS Discussions
Convention on Certain Conventional Weapons. Lethal Autonomous
Weapons Systems framework. The Crown's Agents are
non-lethal by charter — any deviation triggers
immediate revocation.
ICCPR Art. 19 + 17
Free expression and privacy. Bound on every Agent: no
surveillance, no analytics, no covert profiling. §4.11
Honesty Obligation enforces this in code.
Accountability Specification
Every Act, Recorded
No Agent operates in shadow. The accountability surface is part of the charter, not an afterthought.
Identification
Every Agent has a charter URN, a public key, and a declared scope.
urn:sovereign:agent:<class>:<instance>.
Action signing
Every action is signed with ML-DSA-65 against the Agent's
charter intermediate. Unsigned acts are not Agent acts.
Registry inclusion
Every action is appended to the public Merkle log within the
Witness Agent's SLA. No off-record acts.
Error attestation
When an Agent malfunctions, it must emit a 660 (Verification
Chain Broken) or 680 (Registry Integrity Failure) attestation
before halting. Silent failure is itself a 604.
Council jurisdiction
All Agent decisions are appealable to the Council under
code 612 (Dispute Resolution Required). Council may overturn,
modify, or affirm.
Revocation
Any Agent in breach of its charter scope is revoked under
code 771. Revocation is immediate; dependent acts are flagged
for re-verification.
Independent audit
Quarterly attestation under code 741. Audit logs published
in full. External parties may run their own Verification
Agent against the registry to independently confirm
conformance.
Sunset
Every Agent class has a sunset date (default 1 year). Renewal
requires Council statement (732) re-affirming scope and limits
on the basis of experience.
Implementation Phase
What is Live Now
The full Agent fleet ships across Phase 2 and Phase 3. The first two Agent classes are live in Phase 1.
Phase 1 — Live
Class α Verification Agent at
/api/verify.
Class ζ Witness Agent publishing to /.well-known/sovereign/.
Both signed under the Crown Charter v1.
Phase 2 — Next
Class β Triage, Class γ Compliance, Class δ
Settlement go online with
/api/charter,
/api/certify, /api/bank/issue.
Phase 3 — Then
Class ε Sentinel (FODP-v1), Class η Translator,
Class θ Treaty-Watch on full Merkle registry with STH.
Phase 4+ — Continuous
New Agent classes added by amendment under code 701/702.
Conformance suite open to outside verifiers.