From lex ferenda to lex lata? AI and international law: converging existing principles with emerging standards

The rapid advancement of artificial intelligence (AI) has outpaced efforts to establish binding global regulation, leaving AI governed by a fragmented patchwork of national laws, regional instruments, and soft-law guidance. This paper advances two connected arguments. First, a complementarity thesis: in the absence of comprehensive AI-specific global standards, the established norms of international human rights law, international humanitarian law, and international criminal law supply a binding, near-universal baseline that fills regulatory gaps, guides interpretation, and disciplines the design of new instruments. Second, and more distinctively, a crystallisation thesis: the striking convergence of soft-law instruments – from the Organisation for Economic Co-operation and Development Principles (OECD) and the United Nations Educational, Scientific and Cultural Organization (UNESCO) Recommendation to consensus General Assembly resolutions and the Council of Europe Framework Convention – around a recurring core of principles, notably transparency, accountability, and human oversight, indicates that general principles of law specific to AI are already forming, even as attempts to conclude binding instruments advance slowly. On this view, a court confronted with an AI dispute need not declare a non liquet: existing international law, thickened by these crystallising principles, precludes a legal vacuum. The paper substantiates these claims by examining the regulatory efforts of the United Nations, OECD, European Union, and standard-setting bodies, the obstacles to global harmonisation, and the application of existing legal principles to AI, arguing that a unified framework should consolidate this emerging normative acquis.