Institutional legitimacy and governance readiness for public safety AI: beyond technical accuracy in government AI systems

Purpose This study aims to examine how institutional legitimacy and governance conditions shape public acceptance of artificial intelligence (AI)-based threat detection systems, demonstrating that technical accuracy is necessary but normatively insufficient for sustainable policy-oriented support. Design/methodology/approach A survey-based research design was used, with data from 510 valid respondents in South Korea. The study applied a two-step structural equation modeling approach comprising confirmatory factor analysis and structural path analysis, alongside bias-corrected bootstrap mediation analysis with 5,000 resamples. Findings Trust in government significantly predicts institutional legitimacy (ß = 0.660, p < 0.001), which, in turn, shapes both performance expectancy and perceived social deterrence. Direct path analysis confirmed 11 of 12 hypotheses, and bootstrap analysis verified significant indirect effects for all major antecedents. The model explains 84.8% of variance in behavioral intention (R² = 0.848). Ethical concern did not directly undermine legitimacy, indicating conditional rather than automatic normative resistance. Research limitations/implications This research advances public-sector AI governance theory by positioning institutional legitimacy as a mediating filter and introducing perceived social deterrence as a policy-relevant cognitive mediator. It proves that governance readiness – not technical accuracy alone – determines durable policy support. Practically, it highlights the need for transparent oversight to secure public authorization. A limitation is its reliance on a South Korean general public sample, warranting future cross-national, multi-stakeholder comparative research. Practical implications Policymakers should complement technical development with governance mechanisms such as transparency, accountability and procedural safeguards. Social implications The study underscores the societal importance of legitimacy-based governance when deploying high-risk AI systems affecting public safety. Originality/value This research advances public-sector AI governance theory by structurally positioning institutional legitimacy as a mediating evaluative filter between normative antecedents and cognitive expectations. It introduces perceived social deterrence as a policy-relevant cognitive mediator and provides empirical evidence that governance readiness – not technical accuracy alone – determines whether AI systems receive durable policy-oriented authorization.