Legal Means of Providing the Principle of Transparency of the Artificial Intelligence

Objective: to analyze the current technological and legal theories in order to define the content of the transparency principle of the artificial intelligence functioning from the viewpoint of legal regulation, choice of applicable means of legal regulation, and establishing objective limits to legal intervention into the technological sphere through regulatory impact. Methods: the methodological basis of the research is the set of general scientific (analysis, synthesis, induction, deduction) and specific legal (historical-legal, formal-legal, comparative-legal) methods of scientific cognition. Results: the author critically analyzed the norms and proposals for normative formalization of the artificial intelligence transparency principle from the viewpoint of impossibility to obtain the full technological transparency of artificial intelligence. It is proposed to discuss the variants of managing algorithmic transparency and accountability based on the analysis of social, technical and regulatory problems created by algorithmic systems of artificial intelligence. It is proved that transparency is an indispensible condition to recognize artificial intelligence as trustworthy. It is proved that transparency and explainability of the artificial intelligence technology is essential not only for personal data protection, but also in other situations of automated data processing, when, in order to make a decision, the technological data lacking in the input information are taken from open sources, including those not having the status of a personal data storage. It is proposed to legislatively stipulate the obligatory audit and to introduce a standard, stipulating a compromise between the technology abilities and advantages, accuracy and explainability of its result, and the rights of the participants of civil relations. Introduction of certification of the artificial intelligence models, obligatory for application, will solve the issues of liability of the subjects obliged to apply such systems. In the context of professional liability of professional subjects, such as doctors, militants, or corporate executives of a juridical person, it is necessary to restrict the obligatory application of artificial intelligence if sufficient transparency is not provided. Scientific novelty: the interdisciplinary character of the research allowed revealing the impossibility and groundlessness of the requirements to completely disclose the source code or architecture of the artificial intelligence models. The principle of artificial intelligence transparency may be satisfied through elaboration and provision of the right of the data subject and the subject, to whom the decision made as a result of automated data processing is addressed, to reject using automated data processing in decision-making, and the right to object to the decisions made in such a way. Practical significance: is due to the actual absence of sufficient regulation of the principle of transparency of artificial intelligence and results of its functioning, as well as the content and features of the implementation of the right to explanation the right to objection of the decision subject. The most fruitful way to establish trust towards artificial intelligence is to recognize this technology as a part of a complex sociotechnical system, which mediates trust, and to improve the reliability of these systems. The main provisions and conclusions of the research can be used to improve the legal mechanism of providing transparency of the artificial intelligence models applied in state governance and business.