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 analysis of high-risk AI compliance under the European Union Artificial Intelligence Act. Its central claim is that AI governance can become a form of oversight theater when compliance is reduced to benchmarks, static documentation, logging, technical narratives, and formal human oversight while real correction remains unavailable in deployment. The paper does not treat Structural Intelligence as a substitute for legal compliance, harmonised standards, notified-body review, or conformity assessment. Instead, it proposes SI as a complementary diagnostic layer for testing whether high-risk AI systems remain answerable under operational pressure. The paper maps the SI Minimal Runtime Sheet onto central AI Act requirements concerning risk management, record-keeping, human oversight, transparency, robustness, technical documentation, and conformity assessment. It argues that human oversight should be evaluated not only by the presence of a human interface, but by real interrupt power, contextual awareness, low-friction contestability, and consequence-bearing authority. It also treats record-keeping as a trace problem, introducing trace laundering as a failure mode in which logs preserve institutional coherence while smoothing away the original friction. Finally, it links risk management to burden-path audit, asking who absorbs the maintenance bill when an AI system is wrong. The paper contributes to philosophy of artificial intelligence, AI ethics, philosophy of technology, and regulatory governance by arguing that trustworthy AI systems should not be judged only by compliance coherence, but by whether they can still be corrected when reality pushes back.