This preprint develops a structural theory of AI governance centered on the concept of "decision without deciders." Rather than treating AI accountability as a problem of regulatory design, policy implementation, or institutional responsibility, the paper analyzes how accountability itself is being infrastructuralized across contemporary socio-technical systems. It argues that responsibility is no longer located in identifiable human actors, institutions, or decision-makers, but is increasingly embedded in distributed architectures of automation, prediction, and delegation. The paper constructs a conceptual framework for understanding AI governance as a transformation of accountability structures, not as a governance failure, but as a systemic reconfiguration of decision-making itself. This work is positioned as theoretical and analytical research. It does not propose policy design, regulatory frameworks, governance manuals, or implementation models. It is intended as a conceptual foundation for interdisciplinary research collaboration, not institutional deployment. Part of the TLIM Research Program (Tri-Layer Integrated Model).This paper contributes to the integrated framework linking predictive classification, causation, and semantic governance. Part of the TLIM Research Program (Tri-Layer Integrated Model). For the foundational definition of the TLIM framework see:https://doi.org/10.5281/zenodo.18667835 Program website:https://tlimresearch.org
H. Tamba (Wed,) studied this question.