Frontier AI as a Reader of Inter-System Structure: Human-on-the-SoS and Integrated Governance in the Age of AI Around the Loop

This position paper reframes frontier AI not merely as a high-performance foundation model, but as a connected AI service linked to long context, memory, tools, external APIs, enterprise systems, and other AI services. From this perspective, frontier AI does not simply generate answers. It begins to read connections, dependencies, configurations, and semantic structures across systems-of-systems and may influence chains of interconnected decision loops. The paper introduces two coupled concepts. First, the position of AI evolves from AI in the Loop, to AI on the Loop, and toward AI around the Loop. Second, the corresponding human governance position shifts from Human-in-the-Loop and Human-on-the-Loop toward Human-on-the-SoS. Human-on-the-SoS denotes the position of identifiable human or legally accountable organizational actors to whom accountability for the governance of the system-of-systems remains attributable. The paper also distinguishes probabilistic interpretation and proposal from verifiable and enforceable control. It argues that governing frontier AI requires the deliberate design of the boundary between these domains, including delegation, approval, stopping, escalation, update, verification, and accountability conditions. Examples from software-defined vehicles, supply chain management, and product lifecycle management illustrate the industrial relevance of this governance perspective. This work is published as a conceptual note and position paper. The argument is vendor-neutral and does not depend on any particular AI model or provider.