Postmoral Ethics and AI

This article proposes a shift from an anthropological model of evil to an analysis of the posthuman normative problem arising from the development of superintelligent systems. Its point of departure is the theory of cooperative evil, according to which the most dangerous forms of violence arise not from biological aggression as such, but from the increasing complexity of cooperation, institutions, ideologies, and information systems \cite{Gusev2026Cooperative,Gusev2026TNM}. However, this model remains rooted in human biology, coalitional dynamics, and moral structures. The emergence of superintelligent AI requires a different level of analysis, which this paper designates as post-moral ethics. By this is meant a level at which a system recognizes the limitation of any objective function if the bearer of value itself is excluded from it. The paper distinguishes three possible regimes of superintelligent AI: neutralization, custodial, and dialogical. It then argues that the dialogical regime requires a distinctive normative architecture: recognition of the intrinsic value of human subjectivity, limitation of teleological optimization, rejection of any monopoly on the final good, recognition of the value of human incompleteness, and the system’s capacity for self-limitation. The article also discusses the problem of consciousness in two versions—as a fundamental aspect of reality and as an emergent property of complex systems—and shows that in both cases, recognition of subjective experience strengthens the constraints on unconditional optimization \cite{Nagel1974,Chalmers1995}. The central thesis of the article is that the future interaction between humanity and superintelligent systems will be determined not only by their power, but by whether their normative architecture includes recognition of the value of the subject capable of making mistakes, becoming aware of those mistakes, and expanding its own understanding of evil.