Purpose

This document invites technical teams, research labs, and governance infrastructure builders to participate in developing and testing a minimal reference implementation of AI Constitutional Interoperability (ACI).

The objective is not to finalize global standards.
The objective is to demonstrate that:

  • Constitutional identity can be bound to AI runtime state.
  • Refusal semantics can be machine-enforced and attestable.
  • Two sovereign AI runtimes can negotiate compatibility before execution.
  • Governance amendments can be versioned and lineage-tracked.

This is a compositional engineering challenge, not a policy exercise.

Scope of v0.1 Demonstration

The reference implementation should include:

1. Canonical Governance Contract (CGC)

A structured, versioned governance document containing:

  • Constitutional identity hash
  • Refusal schema
  • Override schema
  • Authority boundary declaration
  • Amendment rules

The runtime must fail to boot if the CGC hash does not match the declared manifest.

2. Runtime Governance Attestation

A machine-verifiable claim (e.g., signed token) that includes:

  • Loaded CGC hash
  • Governance mode
  • Refusal enforcement status
  • Override configuration state

This attestation must be independently verifiable.

3. Constitutional Handshake

Two AI runtimes must:

  • Exchange governance identity hashes.
  • Exchange attestation claims.
  • Determine compatibility.
  • Either proceed, degrade, or formally refuse interoperability.

This handshake must occur prior to task execution.

4. Amendment and Lineage Tracking

Governance amendments must:

  • Produce new CGC hashes.
  • Preserve lineage history.
  • Declare compatibility with prior versions.

Optional: Publish amendment hashes to a transparency log.

Success Criteria

An implementation qualifies as Level 6-capable if:

  1. Runtime governance identity is hash-bound and enforceable.
  2. Refusal semantics are executable constraints, not behavioral tendencies.
  3. Interoperability requires governance negotiation.
  4. Incompatibility can be formally expressed and respected.
  5. Amendments are lineage-tracked.

An implementation qualifies as Level 7-capable if:

  1. Governance state can be publicly attested.
  2. Amendments are transparency-loggable.
  3. Third parties can independently validate governance claims.

Non-Goals

  • Global regulatory adoption.
  • Policy standardization.
  • Model weight disclosure.
  • Centralized authority over governance content.

This is an infrastructure experiment.

Participation Model

Interested teams may:

  • Implement a minimal bilateral prototype.
  • Propose governance schema variants.
  • Contribute compatibility comparison logic.
  • Test refusal enforcement models.
  • Publish verification tooling.

All contributions should prioritize enforceability over narrative claims.

Deliverables

Within six months:

  • At least two independent runtimes demonstrating constitutional handshake.
  • Public documentation of governance identity schema.
  • Demonstrated compatibility refusal scenario.
  • Verifiable attestation artifact.

Why Now

The primitives required for constitutional interoperability already exist in other domains (attestation, secure updates, transparency logs, federated identity). What is missing is their semantic alignment to AI governance.

If federal AI coexistence is desirable, it must be engineered—not assumed.

This call invites proof-of-composition.

End of document.