Most governance frameworks for agentic AI systems implicitly assume a single agent or a linear chain of agents with a defined workflow. Under those assumptions, governing the output appears sufficient. This paper argues that multi-agent topology — branching, merging, recursion, delegation depth, and re-entry — destroys that assumption entirely. Topology introduces three unavoidable ambiguities — identity, causality, and privilege — that make output-layer governance mathematically incapable of producing admissible governance guarantees. The paper establishes that topology is not a detail but a force multiplier that turns substrate-layer governance from a design preference into a structural requirement. It concludes that any governance standard that does not define substrate-layer primitives is structurally incomplete for multi-agent environments, which now describes the majority of production agentic deployments.
Narnaiezzsshaa Truong (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: