Physical Intelligence Risk & Emergence Framework (PIRE-F)

Artificial intelligence is increasingly embedded in physical systems—robotics, transport, critical infrastructure, and medical devices—where it acts upon matter and energy, producing risks characterized by irreversibility, cascading failure modes, and emergent autonomy. Existing governance approaches, including the NIST AI Risk Management Framework (AI RMF 1.0) and ISO/IEC AI risk guidance, provide organizational scaffolding but under-specify the behavioral enforcement layer required for embodied AI deployments. Meanwhile, the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) imposes binding obligations for high-risk AI systems, including technical documentation, logging, and post-market monitoring, without prescribing a concrete mechanism for demonstrating continuous compliance in real physical environments¹²³. This paper proposes the Physical Intelligence Risk & Emergence Framework (PIRE-F), a novel, model-agnostic governance architecture for AI in physical systems that evaluates risk primarily through observable behavior. PIRE-F formalizes six behavioral-physical governance domains—Material Sovereignty, Behavioral Coherence, Physical Alignment, Boundary Integrity, Emergent Agency, and Consequence Probity—aligned with SCAB principles while articulated for safety engineering and regulatory assessment. The framework introduces quantitative metrics for constraint adherence, coherence divergence, reachability-based boundary risk, agency escalation, and probabilistic trace reconstructability, and defines threshold-driven intervention ladders enabling runtime constraint tightening, authority reassertion, and physical isolation. PIRE-F is shown to instantiate the NIST AI RMF’s Govern–Map–Measure–Manage functions for embodied systems, integrate with ISO/IEC AI risk management processes, and support EU AI Act technical documentation and post-market monitoring through structured behavioral evidence artifacts.