ENACTING RESPONSIBLE AI IN THE SWEDISH PUBLIC SECTOR

Abstract

Artificial intelligence (AI) is becoming increasingly visible across the Swedish public sector, yet it remains unclear how responsible use is enacted in practice. This thesis examines how AI systems are used across Swedish public-sector organisations and what conditions make responsible use possible. Drawing on Recker et al.’s (2025) digital responsibility framework, organised around three interdependent dimensions: accountability, obligation, and dependability, the study combines a document-based mapping of publicly visible AI systems with six semi-structured expert interviews. The central problem is under what organisational conditions responsible use of AI is enacted, and how this aligns with public-sector expectations of legality, fairness, transparency, and respect for residents’ rights. The findings suggest that AI use is expanding but responsible use remains unevenly developed. Accountability is framed around retained human responsibility, but becomes fragile when employees cannot understand, verify, or explain AI-supported outputs. Obligation involves balancing public-sector benefits against possible harms, while dependability is constrained by competence gaps, infrastructure limitations, vendor dependence, and sovereignty concerns. The study contributes to IS research on digital responsibility by applying Recker et al.’s framework as an empirical lens for responsible AI in a public-sector setting. It shows how the three dimensions become visible in practice and identifies legitimacy and organisational preparedness as empirical conditions that shape whether responsible AI use can be enacted. It also contributes to governance discussions about trustworthy AI in a decentralised welfare state shaped by Nordic values and institutional commitments that set a high bar for how AI systems are expected to serve residents.