AI Regulation in the European Union and the USA

The term artificial intelligence has existed for several decades, but its real boom has been recorded in the last five years due to the development of powerful computing capabilities and advances in the field of algorithms. Historically, artificial intelligence began to take shape in the 1950s, when scientists Alan Turing, John McCarthy and Marvin Minsky laid the scientific and technical foundations of artificial intelligence. The term artificial intelligence was first used by John McCarthy in 1956 at a seminar called the “Dartmouth Summer Research Project on Artificial Intelligence.” When examining this concept, we are confronted with different opinions, due to the absence of a universal legal definition and the existence of a large number of ideas, definitions, approaches and theories that are either too inclusive or too specific for a particular sector. Currently, there is no global legal definition of artificial intelligence, and we encounter definitions that are more applicable to regional or local clusters of needs. Several developed countries are currently not interested in international regulation of artificial intelligence, while their relations regarding the development of new technologies are marked by competition (USA versus China) for leadership in this area. At the European Union level, early soft-law instruments from 2018 onwards provided descriptive definitions of artificial intelligence. In the communications “Artificial Intelligence for Europe” (COM(2018) 237 final) and “Coordinated Plan on Artificial Intelligence” (COM(2018) 795 final), artificial intelligence was described as “systems that display intelligent behaviour by analysing their environment and taking action—with some degree of autonomy—to achieve specific goals”. These descriptions served as policy guidance rather than binding statutory definitions. By contrast, the binding Artificial Intelligence Act of 2024 does not contain a single exhaustive legal definition of “artificial intelligence” expressis verbis; instead, it defines the operational concepts “artificial intelligence system” (Article 3(1)), “general-purpose artificial intelligence model” (Article 3(63)) and “general-purpose artificial intelligence system” (Article 3(66)). The distinction between early advisory descriptions and the later hard-law classification of systems is therefore important for understanding the evolution of EU regulatory language. When it comes to AI legislation, the United States has chosen a fragmented, industry-specific approach. There is no single overarching federal AI law. Rather, limited federal statutes, agency guidelines, executive orders, voluntary frameworks, and an increasing number of state-level laws have all contributed to the evolution of policy.