Institutional Boundary Enforcement: Preserving Institutional Authority in AI-Mediated Information Systems

This repository contains the manuscript, figures, and supplementary materials for Institutional Boundary Enforcement: Preserving Institutional Authority in AI-Mediated Information Systems. Abstract Generative artificial intelligence has intensified a governance problem that predates AI: institutions often retain formal authority while practical execution is distributed across tools, platforms, vendors, intermediaries, and specialized information systems. This article develops Institutional Boundary Enforcement as a theory of how authority-bearing institutions preserve decision authority under AI-mediated and distributed execution. Institutional Boundary Enforcement is defined as the use of formal rules, procedures, authorization structures, and accountability mechanisms to preserve the distinction between authority-bearing and non-authority-bearing actors within institutional decision systems. The analysis advances two boundary-defense strategies. Exclusionary Boundary Defense preserves authority by prohibiting non-authorized actors or systems from participating in authority-sensitive functions. Threshold Boundary Defense preserves authority by permitting participation only when disclosure, attribution, verification, or human-contribution thresholds are satisfied. The article also defines Authority Boundary Collapse as the failure condition in which institutional authorization persists even though boundary rules no longer preserve attributable decision authority. NIH restrictions on generative AI use in peer review illustrate exclusionary defense. U.S. Copyright Office guidance for works containing AI-generated material illustrates threshold defense. The article concludes that durable digital governance requires enforceable boundary rules, not only transparency, explainability, or general human oversight in public information workflows and authority-sensitive institutional decision systems. Status Submitted to Information Polity on 22 June 2026.Manuscript ID: IPO-26-0179.DOI: 10.5281/zenodo.20797519