AICOS® OS: Deterministic Decision Authority as the Missing Operating System for Artificial Intelligence Governance

Artificial intelligence systems have achieved unprecedented levels of predictive and analytical capability. Yet as these systems increasingly influence decisions across finance, public administration, healthcare, and other regulated domains, a foundational problem remains unresolved: the absence of explicit, auditable decision authority. While contemporary AI governance efforts emphasize transparency, explainability, and ethical alignment, they largely fail to address the institutional question that ultimately determines legitimacy—who is authorized to decide. This work introduces AICOS® OS, a governance-first conceptual framework that treats decision authority as an operating system concern rather than a byproduct of intelligence. By formally separating prediction from authorization, AICOS® OS ensures that AI systems may inform decisions without autonomously committing institutions to action. Authority remains human-final, deterministic, and institutionally accountable. Positioned as an open science contribution, this preprint defines a new category in artificial intelligence governance: deterministic decision authority as infrastructure. It examines why existing governance approaches fail, articulates the core principles of AICOS® OS, and explores implications for financial systems, public institutions, and legal accountability—particularly within democratic and regulated environments such as the United States. The central thesis is clear: artificial intelligence can inform decisions, but only legitimate authority can commit institutions to action.