Docking Structural Intelligence into the EU AI Act: Minimal Runtime Sheets, Oversight Theater, and Answerable Control in High-Risk AI Systems

This paper develops a Structural Intelligence (SI) analysis of high-risk AI compliance under the European Union AI Act. It argues that regulatory compliance can become a form of oversight theater when conformity is reduced to benchmarking, static documentation, logging, and formal human oversight while the deployed system still resists correction under real operational pressure. The paper does not present SI as a substitute for legal compliance. Instead, it proposes SI as a complementary diagnostic layer for testing whether a high-risk AI system remains answerable in practice through contact, witness, trace, burden-path visibility, and binding revision. It maps the Minimal Runtime Sheet onto central AI Act requirements concerning risk management, record-keeping, human oversight, documentation, robustness, and conformity assessment, and introduces failure modes such as trace laundering and burden export. Its central claim is that the next stage of AI governance requires not only better documentation, but stronger runtime instruments for detecting when visible oversight has become unanswerable control.