AI Governance · 2026-10-28

JAZB Framework v2.0: Human-Sovereign Authority and Independently Supervised Agent Execution

Zenodo framework publication set (seven phases and release notes)Original paperMarkdown source
AI governanceAI agentsauthoritydelegationauthorizationassurance
Key Insight

Separating autonomous execution from independent enforcement makes organizational authority more testable, but verifiable Law, approvals, and effects do not themselves establish the legitimacy of the sovereign organization or provide external redress.

Review

Michael Costner's seven-phase JAZB Framework v2.0 is a material revision of the human-sovereign Authority architecture reviewed here in v1.1. Version 1.1 made organizational Law establishment, bounded delegation, revocation, separated Internal and Border jurisdiction, independent Actuation, and evidence into explicit control objects. Version 2.0 addresses a more difficult proposition: consequential Task AGENT autonomy can be legitimately permitted when its authorization, supervision, receiving-side enforcement, and actual outcomes remain independently governable. The non-normative release notes document changes from v1.4, not v1.1; claims about intermediate-version novelty therefore cannot be inferred from those notes alone.

The principal architectural change is a separation of Commissioner Law, which bounds authorized business work, from parallel Commandant Law, which governs independent Enforcer observation and intervention. Task and Enforcer AGENTs must have distinct protected principals and control surfaces; the existence of a supervisory AI is not by itself evidence of independence. EDA-selected autonomy profiles determine when a Task may execute without interactive human assent and when preview, multiple approvals, or containment apply. Phase VII ties consequential autonomy to mandatory mediation and evidence rather than to the Task's own declarations. This makes a genuinely falsifiable distinction between permitted agency and self-policing automation, while leaving risk thresholds and institutional decision rights with the Enterprise Domain Authority.

Two further extensions matter. Enterprise AI Customs now applies not only at the actual organizational boundary but at explicitly registered internal data-domain Ports, with a Data Authority Map and source-to-destination assessment. Transfer Points remain separate from Internal behavioral Points, preventing the two jurisdictions from silently acquiring each other's decision rights. Current standing Law and approved transactions receive cryptographic integrity and lifecycle checks; exact immutable change-set manifests, resource preconditions, receiving-side verification, and truthful partial-failure receipts are intended to prevent an authorized intention from being substituted with an unauthorized effect. Phase VII consolidates 295 numbered requirements, 33 beyond the v1.4 identifier range, with version-scoped conformance and evidence obligations.

These are specification-level improvements, not empirical assurance results. Phase VII expressly distinguishes observed effects from simulated or planned outcomes, and the release notes disclaim representative production validation of the new mechanisms. A credible assessment must attempt Task/Enforcer impersonation, direct resource bypass, forged or stale Law, compromised signing keys, internal Port circumvention, batch fragmentation, replay, degraded telemetry, and partial commits. Passing a textual conformance mapping or producing signed receipts cannot establish that every effective write path is mediated. Nor does cryptographic authenticity make a false observation true. Independent tests must establish the causal chain from revocation or denial to the receiving system's actual behavior.

The strongest scope defense is also the correct one: JAZB does not purport to replace public law, privacy controls, external adjudication, or accreditation. Its EDA is the highest active organizational governance authority, not a source of universal legitimacy. That delimitation preserves architectural coherence but leaves the earlier review's central institutional question unresolved. An organization can correctly sign, interpret, enforce, and reconstruct a discriminatory or procedurally defective Law. Internal review of Enforcer findings is not necessarily accessible external contestation by affected persons. Version 2.0 more rigorously governs delegated execution while leaving independent constraints on the sovereign root, capture-resistant Law establishment, and meaningful outside remedy to organizational and legal arrangements beyond its demonstrated conformance boundary.

The next material evidence would be independently reproducible, adversarial cross-component implementations showing effective mediation under compromise, revocation during ongoing tasks, legally constrained policy change, contested evidence and human review, and recoverable outcomes across partial failure. Version 2.0 advances JAZB from a detailed authority taxonomy toward testable execution governance, but its legitimacy and operational-assurance claims must remain bounded to what a deployment can actually demonstrate.

Key Insight

Separating autonomous execution from independent enforcement makes organizational authority more testable, but verifiable Law, approvals, and effects do not themselves establish the legitimacy of the sovereign organization or provide external redress.

Appears in these collections

Continue exploring

Related reviews

More in AI Governance
AI Governance · 2026-10-01

JAZB Framework v1.1: Judiciary AI Zero-Trust Broker

Zenodo framework publication set

JAZB makes machine authority governable by turning provenance, delegation, revocation, jurisdiction, actuation, and evidence into explicit control objects, but organizational authority becomes legitimate governance only when the human sovereign root is itself constrained, contestable, and externally accountable.