Structural Indeterminacy and the Limits of Legal Attribution in AI Systems: A Translation Between Formal Impossibility and Legal-Facing Governance Failure

Recent formal work has established that single-locus accountability frameworks fail structurally — not contingently — once the compound autonomy of human-agent collectives exceeds a computable threshold. This result, derived independently by Tibebu (2026) and the author through a parallel research program, transforms what was previously understood as a practical governance challenge into a proven impossibility. This paper builds a translation between that formal result and the empirical record of legal proceedings involving AI systems. Across three domains — criminal risk assessment, algorithmic discrimination, and AI-assisted medical decision-making — courts and institutions have repeatedly encountered the same underlying structural condition: decisions are required in the presence of signals whose causal attribution cannot be established within the admissible observation regime. These are not failures of transparency, documentation, or oversight design. They are instances of structural indeterminacy. The paper then characterizes the class of governance architecture the impossibility result demands: one whose organizing principle is the refusal to form authority where attribution cannot be structurally supported. A concrete instantiation of that class is described.