Toward Meaningful Mortality in AI Systems: Intrinsic Degradation, Identity, and Succession

Artificial systems are naturally persistent in ways biological individuals are not: theycan be copied, restored, migrated, retrained, and replaced across instances. This createsa gap between internal degradation and any stronger notion of death. In this paper, westudy that gap through a proof of concept of intrinsic mortality in a neural model withcumulative endogenous degradation. The model exhibits a long stable phase followed byrapid terminal collapse, showing that model-internal mortality is feasible at the instancelevel. We then validate, empirically and conceptually, that this mortality is insufficient atthe system level: restore, cloning, and code bypass preserve useful operation beyond thedeath of any single execution trajectory. We argue that the problem must be analyzedthrough a triad: mortality, identity, and succession. Mortality concerns whether an instancecan terminate irreversibly; identity concerns what persists as the same entity; successionconcerns continuation of function across distinct instances. The contribution is not a solveddesign, but a clearer problem definition supported by both proof-of-concept and attackexperiments. Under standard software assumptions, meaningful mortality is difficult becausepersistence is preserved by copying, modification, and replacement.