The Legal Fiction of Control: A Doctrinal Critique of the "Standard of Care" Mandated by Article 14 of the EU AI Act

Abstract

Article 14 of the Artificial Intelligence Act (Regulation (EU) 2024/1689) mandates a "human-in-the-loop" framework to monitor, detect, and override high-risk AI operations. This mechanism promises semantic safety by anchoring automated decision-making to accountable human judgment. Utilizing doctrinal legal research, this paper exposes a critical structural failure within this paradigm: the systemic conflation of the provider’s ex-ante design duties under Article 14 with the deployer’s ex-post operational mandates under Article 26. Because high-risk systems leverage complex deep-learning architectures, their mathematical execution remains epistemically opaque, presenting an insurmountable "black-box" problem that real-time interfaces cannot resolve. By failing to provide a formal statutory definition of explainability, the Act demands cognitive and temporal standards of care that require the performatively impossible, violating the maxim lex non cogit ad impossibilia. Furthermore, this paper integrates a private law critique demonstrating how the qualitative nature of Article 14 enables technology providers to weaponize compliance as a civil liability shield under the Product Liability Directive (Directive (EU) 2024/2853), displacing strict product risks onto downstream deployers. Drawing analogies from CJEU and ECtHR jurisprudence, the analysis demonstrates that Article 14 constructs a dangerous legal fiction. It concludes by presenting two remedial pathways: a narrow judicial reinterpretation of oversight as an ex-post audit framework, or a broad legislative pivot toward a provider-centric strict liability regime.