Where the Frontier Stops: A Response to "Agents, Orchestration, and the EU AI Act" (SSRN 6509259) and the Substrate Gap It Does Not Close

This note responds to "Agents, Orchestration, and the EU AI Act" (SSRN Abstract ID: 6509259), a recent legal-technical paper that represents one of the more technically literate and legally grounded attempts to address agentic AI governance within the EU regulatory framework. The paper advances the conversation in meaningful ways — it is more orchestration-aware, more autonomy-specific, and more substrate-adjacent than most of the current governance literature. This commentary acknowledges those advances and identifies three structural gaps where the paper stops short of the deeper architectural reframing that agentic systems require: its assumption that the EU AI Act can be adapted to agents when the Act's foundational categories are structurally misaligned with agentic systems; its treatment of orchestration as a governancetarget rather than a governance substrate; and its absence of deletion-oriented architecture — the reversibility, state rollback, and lineage integrity primitives that accountability without undoability cannot provide. The APR-Series substrate governance framework, developed and timestamped independently at Soft Armor Labs beginning in 2025, provides the architectural foundation that the papergestures toward but does not reach. The paper stops at the frontier. This note names what lies beyond it.