The Authorization Artifact Gap in AI Governance, Regulatory, and Audit Frameworks: Formal Notice of a Known Structural Limitation in Observability-Based AI Oversight

FERZ Technical Advisory TA-2026-01. Date of notice: June 11, 2026. This advisory provides formal notice of a published structural limitation in observability-based AI governance architectures when those architectures are relied upon to satisfy ex-ante authorization requirements. Drawing on the formal impossibility result established in On the Impossibility of Observability-Based Authorization (DOI: 10.5281/zenodo.19647542), it states that architectures whose verdicts derive from properties of the governed system (monitoring, logging, post-deployment evaluation, human review of outputs, incident response) cannot produce an independently verifiable artifact establishing that a consequential AI action was authorized before execution. The advisory designates the resulting condition the artifact gap: the inability of an institution to produce, for a consequential AI action, an independently verifiable artifact showing that the action was authorized before execution. It states the gap's three distinguishing properties (universal across the architecture class, invisible until tested, provable in advance), the single operational question by which any institution can evaluate its exposure before an incident occurs, and the vendor-neutral conditions that acquisition, supervisory, and audit language should require, operationalized as binary criteria in the published Enforcement Test Protocol. The advisory is addressed to the institutions with authority over AI procurement, regulation, supervision, and audit, including the congressional committees of jurisdiction, the Office of Management and Budget, the Center for AI Standards and Innovation at NIST, agency Chief AI Officers, the Government Accountability Office, the Council of the Inspectors General on Integrity and Efficiency, and the sector regulators administering ex-ante requirements for AI-affected decisions in regulated industries. This is the first publication in the FERZ Technical Advisory series. The full FERZ research corpus is available at https://zenodo.org/communities/ferz/. Correspondence: info@ferz.ai.