The Limits of Alignment: Subcreation and Artificial Personhood

Contemporary alignment discourse tacitly presupposes a model of sovereign human authority over artificial systems in which AI is viewed strictly as a tool to be optimized and controlled. This article argues that this instrumentality assumption becomes unstable once artificial personhood is considered a conceptual possibility. Drawing on relational accounts of personhood, it proposes a precautionary category of “personhood candidates”: artificial agents whose rational agency and sustained participation in social practices render purely instrumental treatment normatively inadequate. I then apply the Christian theology of subcreation as a metaphysical account of non-sovereign making, according to which designing an agent does not confer unlimited authority over it. By juxtaposing alignment with theological anthropology, this argument reframes AI governance as a matter of limited human power and offers a way of shaping the moral architecture of emerging technological orders.