Правові механізми аудиту та підзвітності в системах штучного інтелекту в контексті міжнародного права
Annotatsiya
This article is devoted to the oversight of AI systems’ activities and the provision of accountability obligations within the framework of international legal norms. The study provides a comparative analysis of existing regulatory mechanisms across various legal systems, examining the legal and normative challenges they present during implementation. The article discusses the oversight and accountability mechanisms of AI systems based on key international legal sources, including the European Union Artificial Intelligence Act, the UNESCO Recommendation on Ethics in Artificial Intelligence, the Organisation for Economic Co-operation and Development (OECD) Principles on Artificial Intelligence, and Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law (draft). The research examines the methodological foundations of technological and process-oriented types of IT audits, including algorithmic, performance, security, and ethical audits. It also clarifies the responsibilities of various participants in the accountability chain (developers, providers, users, and regulators). The article highlights the practical implications of technological challenges such as the “black box” nature of IT systems, their dynamic and adaptable nature, and information bias, as well as regulatory gaps such as inter-jurisdictional coordination problems and uncertainty about legal liability. In this context, a comparative analysis is conducted of different regulatory practices, such as the European Union’s risk-based model, the United States’ sector-specific approach, and China’s state control model. The analysis demonstrates that these different approaches hinder the establishment of uniform global standards and create legal uncertainty for companies operating across borders. The role of international standards and certification mechanisms, as well as the importance of ensuring human control and transparency in addressing these challenges, is emphasized. The research findings demonstrate that effective auditing and accountability of IT systems within the context of international law requires integrating technology-focused and process-focused approaches and strengthening international legal cooperation. The article offers specific recommendations to address existing regulatory gaps, including the adoption of a national legal framework, the implementation of mandatory human rights impact assessments, strengthening transparency and accountability obligations, ensuring human oversight, and the establishment of independent oversight bodies. These steps will ensure the fair and safe development and implementation of AI technologies that respect human rights.