Artificial intelligence and autonomous robotics are entering hospitals. This is no longer a question of whether but of when, how fast, and — most critically — under whose governance. This paper proposes SMART-H 3.0 (Smart Medical Autonomous Robotics Technology for Hospitals), a governance-native architecture for AI-enabled hospitals that places infection prevention for immunocompromised patients as its primary constitutional objective. The architecture builds upon two prior published frameworks by the present author: the RAH-SPINE (Recursive Agentic-Human Governance Spine) twelve-layer governance stack, which provides the foundational layered architecture for embedding governance rules into autonomous AI systems; and the Computable Governance Notation (CGN) formalism, which enables machine-readable regulatory policies to compile into executable operational constraints across diverse jurisdictions. Together, these frameworks serve as the proposed central nervous system of the hospital-as-organism. At the heart of the architecture lies a six-phase recursive governance loop — Sense, Score, Simulate, Act, Review, Learn — through which every autonomous decision is continuously evaluated, validated against constitutional constraints, and subject to medical-led human review. The paper introduces the Constitutional Autonomous Hospital Spine (CAH-SPINE), a fourteen-layer governance stack extending RAH-SPINE with hospital-specific components including digital twin simulation, autonomous swarm robotics coordination, sovereign AI processing, and post-quantum cryptographic identity. A constitutional zone model classifies hospital areas by immunocompromised patient vulnerability, with non-optimisable safety floors that cannot be overridden by AI optimisation. The proposal is explicitly conceptual (TRL 1-2): no prototype exists, no clinical validation has been conducted, and all quantitative parameters are illustrative governance modelling constructs. The paper contributes an architectural hypothesis — that the hospital of 2035 may be defined less by the sophistication of its robots or the power of its AI than by the maturity of its governance.
HORACIO BRIZUELA (Mon,) studied this question.