Risk Assessment for AI Applications in Security Systems: Challenges and Opportunities

Given the practice of the increasingly widespread application of artificial intelligence (AI) in the security system, there is a need to consider the opportunities and challenges of risk assessment in this area. Existing methodologies, which are applied in the Republic of Serbia, such as international and national standards ISO 31000, EN 17640 and SRPS AL.2.003:2025, provide general frameworks for risk assessment, but do not include the specifics that the application of AI technologies in security systems entails. The autonomy, unpredictability and complexity of AI models, as well as the ethical and legal aspects of their application, require finding a new approach that will enable a comprehensive risk assessment. The paper points out the shortcomings of the existing regulatory framework in this area, suggests supplementing the existing risk assessment models and finding a new, comprehensive methodology for risk assessment in the application of AI technologies with an emphasis on the security system. The paper also provides recommendations for improving the regulatory and methodological framework in accordance with the EU AI Act, including amendments to national standards and the development of sectoral guidelines. It is concluded that systemic adaptation to risk management is necessary for the responsible and safe application of AI technologies in security systems.