From Democracies to Autocracies: How AI Systems Enable Authoritarianism by Design

The paper investigates the lifecycle of six AI systems across various political regimes, revealing how features such as centralized data utilization for law enforcement, regulatory gaps, and weak human oversight contribute to authoritarianism. It highlights that both centralized and fragmented AI systems can enable authoritarian practices, depending on governance structures, and emphasizes the need for developers and policymakers to address these design and operational risks. This research is crucial for AI practitioners to understand the implications of their systems in different political contexts and to implement safeguards against misuse.