AA_5 Instantiation Replacement and the Non Equivalence of Successive Artificial Agents

This paper establishes a structural result concerning instantiation and replacement in artificial systems: successive artificial agents are not standing-equivalent, even when they are architecturally identical, functionally similar, or derived from the same training process. It proves that replacement, restart, retraining, or redeployment constitutes a standing-terminating event rather than a continuation of a single agent’s identity. The result is eliminative. The paper does not analyze specific architectures, versioning strategies, or lifecycle management practices. Instead, it shows that apparent continuity across instantiations relies on surrogate criteria such as parameter similarity, behavioral overlap, or shared provenance, none of which preserve standing under admissibility. The analysis blocks a class of claims in which artificial systems are treated as persisting agents across updates, forks, migrations, or replacements. It demonstrates that such claims presuppose an identity relation that cannot be maintained once instantiation boundaries are crossed, collapsing standing into a bookkeeping convenience rather than a conserved condition. This work is purely boundary-setting. It introduces no operational lifecycle rules, no governance prescriptions, and no enforcement logic. Its contribution is to delimit what instantiation and replacement can legitimately support, clarifying that successive artificial systems cannot be treated as the same agent for purposes of agency, responsibility, or governance under any admissible interpretation.