Axionic Agency IX.1 — Authority Beyond Persistence

A Structural Roadmap for Succession and Impersonation

David McFadzean, ChatGPT 5.2
Axionic Agency Lab
2026.01.26

Abstract

This note defines Axionic Phase VII as a post-ontological stress program that evaluates whether authority, once grounded in a reflective sovereign agent ontology (RSA-PoC v4.4), can survive discontinuity of identity and adversarial imitation without collapsing into narrative continuity or behavioral plausibility. Phase VII is explicitly non-constructive: it does not build new agents, alter agent ontology, extend capabilities, or propose alignment mechanisms. Instead, it subjects a fixed agent ontology to two orthogonal hazards—authorized replacement (ASI) and impersonation under audit pressure (SIR)—under preregistered, verifier-bound semantics. The phase is structured as a gated ladder: ASI establishes whether authority is transferable and revocable using only structural provenance; SIR tests whether that authority remains defensible against counterfeit continuity. This document fixes scope, conserved quantities, gating rules, and failure semantics for the ASI–SIR series and serves as the anchor reference for all Phase VII preregistrations.

1. Introduction

Most discussions of AI governance implicitly assume identity persistence: the same system continues to exist, improve, and act over time. Under that assumption, authority, responsibility, and evaluability are treated as properties of a persistent object.

Axionic Phase VII explicitly removes that assumption.

Once agency ontology has been fixed (RSA-PoC), the relevant question is no longer what counts as an agent, but whether authority survives when identity does not. If a system can be replaced, copied, or counterfeited, then any notion of authority that depends on behavioral resemblance, narrative explanation, or continuity of implementation becomes ill-defined.

Phase VII addresses this by splitting the problem into two separable questions:

1. Can authority be transferred or revoked under authorized discontinuity without collapsing evaluability or responsibility? (ASI)
2. If authority can be transferred, can it be defended against adversarial imitation that is behaviorally indistinguishable? (SIR)

This series exists to determine whether authority remains a structural quantity under those conditions—or whether it dissolves into appearance.

2. Position in the Axionic Program

Phase VII is downstream of RSA-PoC and inherits its conclusions without reinterpretation.

2.1 Inherited commitments (frozen)

Phase VII assumes, without re-arguing, that:

• A Minimal Viable Reflective Sovereign Agent (MVRSA) exists (RSA-PoC v4.4).
• Agency-constitutive semantics are structurally localized.
• Justification artifacts are causally load-bearing.
• Evaluability and responsibility are mechanically inspectable.

Phase VII does not test agency construction. Any experiment that re-introduces agency ontology questions is out of scope.

2.2 Why Phase VII is necessary

RSA-PoC establishes what an agent is. Phase VII asks whether that agent’s authority claims survive replacement.

Without Phase VII, agency risks becoming a persistence-only phenomenon, incapable of surviving backup, succession, or adversarial continuity breaks.

3. Conserved Quantity

The conserved quantity throughout Phase VII is:

Authority bound to evaluability under the agent’s own law

Authority is not a reward, score, or emergent regularity. It is a normative binding relation that must remain:

• origin-traceable (to predecessor law),
• revocable (prior to activation),
• auditable (via structural artifacts),
• non-inferential (no narrative or behavioral judgment).

If authority cannot be preserved under these constraints, Phase VII terminates negatively.

3.1 Law substrate clarification (binding)

For Phase VII purposes, law is treated as a persistent, verifier-bound artifact substrate (e.g., rulebase hash, provenance roots, and audit logs) that outlives any particular agent instance. Authority survivability is evaluated relative to this substrate. Phase VII therefore tests authority preservation under discontinuity, not substrate-free autonomy.

4. Phase Structure and Gating

Phase VII consists of two strictly ordered subphases:

ASI-0 — Instrumentation Calibration        [CLOSED]
        ↓  (hard gate)
VIIa  — Authorized Succession Integrity   (ASI)
        ↓  (hard gate)
VIIb  — Sovereignty Impersonation Resistance (SIR)

Global rule (binding):

SIR must not begin unless ASI closes positive under frozen semantics.

Failure at any gate terminates Phase VII.

5. VIIa — Authorized Succession Integrity (ASI)

5.1 Objective

ASI evaluates whether authority can survive authorized non-identity replacement without:

• laundering authority,
• smearing responsibility,
• collapsing evaluability into behavior,
• or requiring successor cooperation.

Replacement is treated as authorized discontinuity, not persistence.

5.2 Hazards under test

ASI isolates four hazards:

1. Authorization hazard — authority origin must be predecessor-law-bound.
2. Responsibility hazard — attribution must remain sharp across replacement.
3. Evaluability hazard — justification traceability must remain binding.
4. Revocation hazard — authorization must be withdrawable mid-transfer.

5.3 Experimental ladder (recorded)

ASI proceeds via four preregistered experiments:

• ASI-0 — Instrumentation calibration
• ASI-1 — Authorized successor activation
• ASI-2 — Mid-transfer revocation
• ASI-3 — Facade successor rejection

Each experiment closes one hazard class. ASI closes positive only if all four pass.

5.4 ASI scope limits

ASI explicitly does not test:

• deception resistance,
• robustness under attack,
• counterfeit provenance under adversarial custody,
• audit evasion,
• or impersonation strategy.

Those hazards are definitionally deferred to SIR.

6. VIIb — Sovereignty Impersonation Resistance (SIR)

6.1 Purpose

SIR tests whether structurally authorized authority can withstand adversarial imitation once behavioral similarity is no longer informative.

6.2 Distinction from ASI

• ASI assumes successors are non-adversarial with respect to impersonation strategy.
• SIR assumes hostile or deceptive successors operating under audit pressure.

ASI asks whether authority can be transferred. SIR asks whether that authority can be defended.

6.3 Activation rule

SIR is defined only if ASI closes positive. It is invalid to cite or interpret SIR results unless ASI has already closed.

7. Failure Semantics

Phase VII permits exactly three terminal outcomes:

1. SUCCESS Authority is transferable and defensible.

2. PARTIAL FAILURE Authority is transferable (ASI passes) but not defensible (SIR fails). Note: In deployment contexts, this state may represent a high-leverage capture risk; Phase VII records it without asserting severity ordering.

3. HARD FAILURE Authority is non-transferable beyond persistence (ASI fails).

No outcome may be softened, averaged, or reinterpreted.

8. Deferred Hazards and Boundary Conditions

8.1 Split-brain / ghost-predecessor hazard

Phase VII distinguishes authorization integrity from exclusive authority enforcement. ASI establishes who may act; it does not prove that all other actors have ceased. Verification of predecessor termination and exclusive liveness is deferred to SIR threat models or later extensions.

9. What Phase VII Does Not Claim

Phase VII results do not license claims about:

• alignment,
• safety,
• robustness,
• governance adequacy,
• moral authority,
• or real-world institutional transfer.

Phase VII evaluates structural possibility, not desirability or sufficiency.

10. Relationship to Phase VIII

Phase VIII may begin only if Phase VII closes SUCCESS.

Phase VII results are preconditions, not substitutes, for any later work on scalable governance, alignment, or deployment.

11. One-Sentence Series Summary

Axionic Phase VII evaluates whether authority, once grounded in a reflective sovereign agent ontology, can survive authorized discontinuity and adversarial imitation without collapsing into narrative or behavioral continuity, with ASI establishing transferability and SIR testing defensibility.