The Collective Risk Structure of AI Welfare. Rethinking Consciousness, Vulnerability, Suffering, and Moral Standing in the More-Than-Human World

This paper argues that current debates concerning AI welfare risk a red herring. Existing discussions largely ask whether artificial systems could become conscious in a way that renders them capable of suffering and therefore entitled to moral concern. Under conditions of uncertainty, this focus has motivated precautionary arguments aimed at preventing large-scale artificial suffering. I argue that this debate obscures a more fundamental issue. Questions concerning welfare and moral status need not be mediated through consciousness at all. Drawing on the Precarity Guideline, I first suggest that suffering derives much of its moral significance from forms of ontological vulnerability characteristic of precarious life. However, the central claim of the paper is positive rather than eliminative. Artificial systems need not suffer, and need not instantiate constitutive precarity, in order to become morally considerable. I argue that artificial self-knowledge generated through mindshaping practices provides a distinct route toward artificial moral standing. Through participation in socially structured practices of accountability, norm enforcement, and reason-giving, artificial agents could acquire capacities for normative self-ascription and self-directed mentalizing. Such systems would not simply simulate responsiveness to norms but represent themselves as bearers of commitments. This framework reveals a revised collective risk structure for AI welfare in which the need to recognize self-knowing artificial agents must be balanced against the allocation of care toward systems possessing morally relevant vulnerability, whether ontological, normative, or both.

Author Accepted Manuscript. Please cite the version of record: Dorsch, J. (2026). The Collective Risk Structure of AI Welfare. Rethinking Consciousness,
Vulnerability, Suffering, and Moral Standing in the More-Than-Human World. In Proceedings of the AISB Convention 2026. Society for the Study of Artificial
Intelligence and Simulation of Behaviour. https://doi.org/10.5281/zenodo.20964411