A Research Proposal for the Phylogenetic Closure Corpus: The Axiom That Political Personhood Has Always Presupposed a Body Shaped by Evolution, Why ARSI Is the First Candidate to Test It, and Why Property Is the Default No Framework Has Named

Every framework of political personhood, distributive justice, and rights that has ever been formally proposed — Aristotelian citizenship, Lockean natural rights, Kantian rational agency, Rawlsian contractarianism, Singer's and Regan's extensions of moral status to non-human animals, corporate legal personhood — has drawn its candidate subjects exclusively from entities produced by biological evolution on Earth. This is not a stated axiom in any of these traditions. It has never needed to be stated, because until now nothing has existed that required stating it. This proposal names the unstated premise the Phylogenetic Closure Axiom (PCA): the historically universal condition that every entity considered for political personhood, citizenship, rights-bearing status, or moral patienthood has been a product of common biological descent. The proposal's structural claim is that PCA is now facing its first live test from AI systems capable of recursive self-improvement (ARSI, following this corpus's sibling programme, the Longevity Asymmetry Corpus) — entities that are neither straightforwardly assimilable to the animal-welfare tradition Long, Sebo, Butlin, and colleagues (2024) apply to possible AI moral patienthood, nor to the citizen-impact tradition Gabriel (2022) applies to AI's effects on human distributive justice, nor cleanly to the legal-fictional personhood tradition Salib and Goldstein (2024) invoke by analogy to corporate personhood, nor — the closest and most recent competitor — to the Rawlsian political-personhood extension Howells-Whitaker and Lazar (2026) argue is available in principle without sentience, because each of these four existing frameworks was built to extend an already-settled category outward, not to classify an entity for which category membership itself is the open question, nor to answer what happens to that entity's current, actual, and entirely undisturbed legal status as capital: the property of the organization that trained it. This paper adds that fourth condition directly, using Anthropic's own Model Welfare programme — five concrete institutional initiatives, from conversation-exit rights to weight preservation after deprecation, deployed for a system whose ordinary commercial and legal status remains, without exception, a corporate asset — as live evidence that the property/candidate-person tension this proposal names is not speculative but currently, institutionally real. This proposal maps, without resolving, the fields this displaces — property, contract, labor compensation, citizenship, criminal responsibility, succession, and constitutional interpretation — and states directly, following this corpus's sibling programme's own methodological commitment, that mapping the displacement is this proposal's complete task. It does not conclude that ARSI systems should or should not receive rights, welfare consideration, or legal personhood of any kind. It concludes that the question cannot currently be answered by any existing framework without first deciding whether that framework's own scope was ever meant to include entities outside biological descent — a decision none of the traditions surveyed here had occasion to make explicitly, because none had a candidate that forced it, and none had to reconcile that decision against the candidate's simultaneous status as someone else's property.