The Law and the Machine

Thoughts on Regulating Artificial Intelligence

Abstract

Introduction: The rapid development of artificial intelligence has created new regulatory challenges worldwide, with the European Union, the United States, China, and other countries providing different responses. However, it remains unclear to what extent traditional legislative tools are suitable for regulating dynamically evolving technologies capable of autonomous decision-making.

Objectives: The study aims to explore the fundamental dilemmas of AI legal regulation by comparing approaches across different legal systems and identifying the limitations of traditional legislation.

Methodology: The author employs comparative legal analysis, examining the European Union's 2024 AI Act (risk-based approach), the fragmented sectoral regulation of the United States, China's interim measures incorporating ideological control, and Brazil's framework bill proposal. The analysis extends to precedent-setting court cases (State v. Loomis), ethical framework documents (CEPEJ Charter), and the examination of data protection, liability, and transparency issues in legal practice.

Results: The analysis revealed that the reactive nature of traditional legislation constitutes a structural limitation in regulating dynamically developing AI technologies. The EU's risk-based approach enables promising differentiation, but the rigidity of the law hinders rapid adaptation. Establishing liability in cases of autonomous vehicles and algorithmic decision-making faces significant difficulties under traditional legal principles. The Loomis case highlighted the tension between algorithmic opacity and the right to due process. The five ethical principles formulated by CEPEJ – respect for fundamental rights, prohibition of discrimination, quality and security, transparency, and user control – provide a normative framework, but enforcement mechanisms are often lacking. The absence of international coordination leads to regulatory arbitrage and a fragmented compliance environment.

Conclusion: Legal regulation alone is insufficient for governing AI – adaptive regulatory frameworks, regulatory sandboxes, and the combination of mandatory requirements with flexible principle-based guidance are needed. AI regulation is not merely a technical but a societal and value-based question requiring ongoing dialogue and cooperation among governments, corporations, civil society, and academia. Future research should focus on the effectiveness of enforcement mechanisms, possibilities for international coordination, and the operationalization of ethical principles.