Artificial Intelligence (AI) is rapidly transforming modern society, influencing various fields from daily life to industrial innovation. Large corporations invest heavily in AI research and development, contributing to technological advancement and economic growth. AI also enhances SME and startup competitiveness. Public interest grows as AI integrates into everyday products and services. However, AI's autonomous goal-setting and subgoal derivation capabilities, while offering efficiency and problem-solving advantages, pose challenges in predictability. AI might act differently from human intentions when setting subgoals, potentially leading to unexpected and even dangerous outcomes. This unpredictability is a significant challenge, especially in scenarios involving human safety and crucial decision-making. This study explores the crucial role of international regulation in ensuring the safe and ethical use of AI, focusing on the experience of the IAEA. The primary objective is to analyze the IAEA's regulatory framework as a model for AI governance, examining how it promotes responsible AI use and manages associated risks. It aims to provide international insights into AI regulation and derive guidelines applicable to other regulatory bodies. The study also provides comparative insights into AI regulatory strategies in the EU and US, offering a comprehensive understanding of international AI regulation.

The literature review examines existing works on IAEA regulations, highlighting their scope and evolution in shaping the regulatory environment for nuclear safety and security. It explores how IAEA standards and procedures have provided stringent safety criteria throughout the lifespan of nuclear facilities. The review includes studies on regulatory compliance of nuclear fusion technology, information security systems at nuclear facilities, regulatory inspection of digital instrumentation and control systems, and an assessment of the efficacy of the current international nuclear safety regulatory framework. These studies provide insights into how the IAEA's regulatory framework needs to evolve, particularly regarding new challenges and opportunities associated with the integration of AI technology. Additionally, a comparative analysis of AI regulation from various countries and regions provides broader context for understanding the IAEA's role in the global AI regulatory environment.

This study employs a systematic review of various data sources to examine the applicability of IAEA regulations to AI. Data collection involved a comprehensive review of IAEA regulations, AI technology standards, and associated international regulatory frameworks. The analysis uses a comparative methodology to evaluate the alignment between existing IAEA regulations and AI technology standards, identifying key interactions and differences. Thematic analysis identifies major themes and patterns in IAEA regulations related to AI. A framework was developed to assess AI integration into the IAEA's existing regulatory standards, structured around three key pillars: regulatory fitness, technical integration feasibility, and regulatory efficacy. The methodology acknowledges the challenges of adapting rapidly evolving AI technology to existing regulatory frameworks, such as the need for continual updates and the complexity of AI technology.

The analysis of the IAEA's regulatory practices provides insights into how these could be applied to AI technology in the nuclear field, particularly examining regulatory aspects when AI is utilized in the operation and maintenance of nuclear facilities. The study examines the IAEA's benchmarks regarding risk assessment, accident prevention, and safe management of nuclear facilities with AI-based systems. Recent developments, like the Biden administration's AI executive order, are analyzed for their influence on AI regulation and the IAEA's regulatory approach. The integration of AI into the IAEA framework presents both challenges and opportunities. Challenges include the need for continual adaptation and innovation in response to the rapid advancement of AI technology and the complexity of the technology itself. Opportunities include enhanced monitoring, decision-making processes for nuclear safety and security, and improved risk detection and accident prevention. The study proposes specific implementation measures for AI safety regulations by referencing the IAEA's structure, including standardized safety standards, independent supervisory systems, emergency response systems, and international information-sharing platforms. The roles and responsibilities of various stakeholders—governments, industries, research institutions, international organizations, and civil society—are explored in the context of AI regulation, drawing parallels with the IAEA's approach.

The research findings contribute to understanding the applicability of the IAEA's regulatory framework to AI, particularly AI applications in the nuclear domain. The analysis provides important insights into how the IAEA regulatory framework can accommodate the rapid advancement and integration of AI technology. The study highlights the challenges and opportunities that emerge in adapting the IAEA framework to the evolving AI landscape, including the need for regulatory flexibility and innovation to address the rapid pace of AI technological progress and the complexity of AI systems. The discussion offers practical recommendations for policymakers and practitioners, including the need for flexible approaches, clear standards and guidelines, international cooperation, and stakeholder participation.

This study demonstrates the necessity of safety regulations for AI, drawing valuable lessons from the IAEA's successful nuclear safety regulatory framework. While the IAEA's approach offers valuable insights, directly applying its model to AI faces limitations due to AI's unique characteristics and risks. Future research should focus on developing AI-specific regulatory strategies that address these challenges and foster public trust in AI systems. This requires ongoing collaboration between policymakers, researchers, and stakeholders to adapt and refine regulatory frameworks in light of the ever-evolving AI landscape. The need for an independent regulatory system, flexible regulatory models, harmonization with existing regulations, and timely responses to rapid AI development are emphasized.

The study acknowledges the limitations of directly applying the IAEA's nuclear safety regulatory framework to AI, given the unique characteristics and risks of AI technology. The rapid evolution of AI, its complexity, and broad societal impacts present challenges not fully captured by the IAEA's approach. The societal and ethical implications of AI extend beyond the IAEA's primary focus on nuclear safety and security.