Development of effective methods and tools for the auditing AI algorithms by Supreme Audit Institutions

Abstract

This article proposes an AI Audit Framework for Supreme Audit Institutions, focusing on public sector usage. It addresses the need for transparency, fairness, accountability, and alignment with ethical and legal requirements. The authors discuss the rise of AI, particularly generative AI and large language models, underscore the evolving regulatory environment, and identify a gap in existing AI audit processes. The article draws on international standards and best practices to offer a methodology for auditing AI algorithms across their entire lifecycle, including risk categorization, data governance, and bias assessment. It also details how generative AI poses new challenges that require specialized guidelines. Recommendations highlight interdisciplinary collaboration and continuous skill development to ensure responsible AI governance.