The limits of technical concretization of legal principles in artificial intelligence regulation

Abstract

Introduction. The relevance of the study is determined by the tension between the abstract valuebased content of legal principles and the need to establish verifiable requirements for artificial intelligence (AI) systems. The extent to which the content of a principle may be translated into technical parameters without replacing it with a predetermined rule remains an open question in legal theory.Purpose. To determine the procedure and limits of the technical concretization of legal principles in AI regulation.Theoretical basis. The methodology combines formal legal, comparative legal, systemic structural and formal logical methods.Results. The study distinguishes between substantive principles protecting individual rights and values and safeguard principles ensuring the verifiability and contestability of algorithmic decisions. Technical concretization is defined as the legally mediated translation of particular normative projections of a principle into requirements for the data, procedures and outputs of an AI system. Articles 3 and 64 of the Russian Labour Code are used to identify gaps between the prohibition of discrimination and its technical and procedural enforcement in algorithmic recruitment. A typology of the results of concretization according to their degree of technical formalizability is proposed. Structural-logical, axiological and epistemic-technological limits are identified. Statistical disparity between groups is shown to constitute not proof of discrimination but grounds for a rebuttable presumption and a duty to disclose selection criteria. An institutional mechanism is proposed that includes shifting the burden of proof, adversarial examination of expert evidence and model maps for selecting fairness metrics. The article specifies which elements may be tested within experimental legal regimes and which require amendments to labour and procedural legislation.Conclusions. Technical concretization can cover only particular normative projections of a principle, while the final legal assessment of competing values in legally significant decisions requires meaningful human oversight.