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

Artificial intelligence systems can enable authoritarian governance even within democratic states, according to a study that mapped the lifecycles of six AI systems across political regimes from the United States to China [1]. The paper, submitted in 2026 to arXiv, finds that the technical and operational features enabling authoritarianism are not confined to autocracies but are present, in varying configurations, across both democratic and autocratic contexts [1]. The authors drew on academic publications, investigative reports, third-party evaluations, media interviews, and government procurement notices to conduct a systematic, qualitative comparison [2]. Enabling features identified include the centralization and co-optation of administrative data for law enforcement and political punishment, regulatory gaps that fail to deter misuse, weak user compliance that nullifies human oversight mechanisms, and the encoding of protected group traits that identify members of vulnerable populations [2]. The study notes that both centralized and fragmented AI systems can contribute to authoritarian outcomes by exploiting governance gaps. Centralized systems directed by executive authorities, particularly within security and military institutions, often lack formal oversight mechanisms, while fragmented systems diffuse accountability between stakeholders, allowing entrenchment [2]. The findings underscore that AI-enabled authoritarianism is distributed, resulting from design and operational choices made by developers, administrators, and users alike [1]. The authors conclude with recommendations for developers and policymakers to mitigate these risks [2]. The research arrives amid broader debates about the political valence of technology. The term "regime" in political science refers to a system of norms, rules, and decision-making procedures that determines access to public office and the extent of power held by officials, with common categorizations including democratic, autocratic, or hybrid forms [4]. The United States, one of the regimes examined in the study, is a federal republic structured around a representative democracy with a constitutional separation of powers among legislative, executive, and judicial branches [5]. Separately, ideological currents such as the Dark Enlightenment—an anti-democratic, anti-egalitarian movement that advocates for authoritarian capitalist city-states—have gained traction in parts of Silicon Valley and among some political figures associated with former U.S. President Donald Trump, including strategist Steve Bannon and Vice President JD Vance [3]. The movement has been described by University of Chichester professor Benjamin Noys as "an acceleration of capitalism to a fascist point" [3]. China, the other pole of the study's regime spectrum, hosts several major AI developers whose models are distributed under open-source licenses. These include Alibaba Cloud's Qwen family [6], Z.ai's GLM models released under the MIT License since July 2025 [7], and DeepSeek, whose R1 model debuted in January 2025 with training costs reported at US$6 million—far below the US$100 million cost for OpenAI's GPT-4 in 2023 [8]. The U.S. Commerce Department blacklisted Z.ai in January 2025, citing national security concerns [7].