THE ETHICS OF ARTIFICIAL INTELLIGENCE: BOUNDARIES AND RESPONSIBILITY

Abstract

Artificial intelligence is increasingly integrating into everyday life; however, its use is accompanied by a significant number of ethical challenges. The lack of algorithmic transparency, potential discrimination, risks associated with military applications, and issues of responsibility for autonomous decisions require thorough analysis. This study examines approaches to AI regulation worldwide, particularly in the European Union, the United States, China, and Ukraine. It has been established that the EU applies strict regulatory frameworks to protect human rights, the US favours self-regulation by AI development companies, China enforces state control over AI development and implementation, which raises concerns regarding the protection of citizens’ rights, while Ukraine is only beginning to establish its own regulatory framework.The research identifies the fundamental ethical principles that should guide AI usage and, based on these, proposes recommendations for regulating the technology. These include the necessity of algorithmic transparency, personal data protection, the right to be forgotten, restrictions on autonomous decision-making in critical areas, and the establishment of international standards. The study analyses the prospects for regulatory development and highlights the importance of balancing technological progress with oversight to prevent the uncontrolled use of AI.Additionally, issues of responsibility for decisions made by artificial intelligence and the necessity of independent oversight over its implementation were considered. The study examines possible scenarios for technological development under global standardisation, as well as the risks that may arise from insufficient regulation. The findings demonstrate that only a comprehensive approach to regulation and accountability will help mitigate potential threats and ensure the safe development of AI for society, minimising the risks associated with its uncontrolled application.