The rapid development of artificial intelligence (AI), particularly in the domains of generative models and large language models (LLMs), has generated intense debate in science and technology studies, media studies, and the broader field of social sciences. While AI has long been discussed as a transformative technological paradigm, its accelerated integration into digital ecosystems since the early 2020s question its implications for illiberalism studies.
This question is especially relevant given the inevitable intersection between AI and digital authoritarianism, a concept that has gained prominence to describe the use of digital technologies to “surveil, repress, and manipulate domestic and foreign populations.” Recent scholarship emphasizes that digital authoritarianism is not limited to fully autocratic regimes but constitutes a broader analytical framework for understanding how states deploy technologically advanced tools—ranging from predictive analytics to social media manipulation—to consolidate political control.
Yet AI also intersects with the emerging notion of digital illiberalism, described by Jasmin Dall’Agnolla as the practices “that restrict individual autonomy, often under the guise of protecting security and public order, without dismantling democratic structures outright,” practices which can “manifest themselves through arbitrary, pervasive, technology-enabled surveillance, data collection, and algorithmic manipulation.” Understanding this convergence requires examining what AI adds to pre-existing authoritarian and illiberal techniques—such as algorithmic surveillance, targeted repression, and population-level disinformation—but also what fundamentally new capabilities it enables. Grasping this convergence therefore requires a careful examination of what AI is (and what it is not) bearing in mind that “artificial intelligence” functions less as a stable analytical category than as a contested label, largely shaped and promoted by the tech industry for commercial purposes.
Defining AI in its Contemporary Form
AI is a fluid and historically contingent category. What has been labeled “AI” has shifted from “expert systems” in the 1980s to deep neural networks in the 2010s and, most recently, to generative AI and LLMs capable of producing text, images, and multimodal outputs.
Image: “Unraveling AI Complexity – A Comparative View of AI, Machine Learning, Deep Learning, and Generative AI” by Lily Popova Zhuhadar licensed under CC BY-SA 4.0.
Yet the term AI today encompasses a wide variety of software systems, many of which differ profoundly in design, functionality, and societal implications. Under this label we find not only generative models such as ChatGPT, Deepseek, or Gemini, but also search engines, recommendation systems, automated decision-making tools, and surveillance software, including facial recognition and large-scale data fusion systems used by firms such as Palantir. What is commonly called “AI” is in fact a specific algorithmic layer embedded within broader software infrastructures, which process large datasets to produce probabilistic outputs. This algorithmic component can be mobilized in radically different contexts—creative content generation, biometric identification, predictive policing, welfare eligibility assessments, or political microtargeting—each raising distinct political and ethical concerns.
What is commonly called “AI” is in fact a specific algorithmic layer embedded within broader software infrastructures, which process large datasets to produce probabilistic outputs.
Contrary to narratives of machine autonomy, contemporary AI systems depend heavily on human labor at multiple stages of development. Large models require extensive human-curated datasets and continuous human oversight to remain functional. As documented by the academic literature, AI development relies on an internationalized and stratified labor force responsible for data labeling and classification, reinforcement learning with human feedback, model calibration, content moderation, trust and safety operations, and post-deployment error correction. This labor is needed because AI Models are trained on vast quantities of publicly available text (or speech) found in Wikipedia, digital archives, news sources, social media posts, and other user-generated content. What appears to be machine-generated knowledge is in reality a statistical distillation of human-produced discourse, partially shaped by the bias, hierarchies, and the racism, homophobia or sexism embedded in these “digital corpora,” unless it is filtered and checked. Much of this labor is outsourced to workers in the Global South, who perform low-paid, repetitive, and sometimes psychologically harmful moderation tasks essential to making models appear “safe,” “non-toxic,” and “intelligent” to end users. Far from being autonomous, AI systems are therefore “dependent assemblages” composed of human judgement and invisible labor.
What appears to be machine-generated knowledge is in reality a statistical distillation of human-produced discourse, partially shaped by the bias, hierarchies, and the racism, homophobia or sexism embedded in these “digital corpora,” unless it is filtered and checked.
A further defining feature of contemporary AI is its reliance on massive computing infrastructures. Recent analyses show that AI development is characterized by an accelerated “race to scale” with the computational power required for state-of-the-art model training increasing roughly to an eight-fold annual increase. Dataset size “has increased at a rate of approximately 2.4× per year.” Training a “frontier model” (i.e a highly capable general-purpose model) can require tens of thousands of kilowatt-hours, and running the model after training demands significant energy when applied at scale. These infrastructural demands mean that only a handful of firms and states possess the computational capacity, energy supply, and capital to develop “frontier AI systems.” AI development is therefore structurally tied to concentrated economic power, specialized hardware supply chains, and political economies of extraction—whether material (rare earth minerals, semiconductor production) or informational (data harvesting).
Implications for the study of Digital Authoritarianism and Illiberalism
Once these defining dimensions of AI are accounted for, it becomes possible to identify, non-exhaustively, several key intersections between AI, authoritarianism, and illiberalism. Existing research on digital authoritarianism demonstrates how technologies such as predictive analytics, algorithmic surveillance, and data-intensive governance can enhance authoritarian control. The rise of AI compounds these dynamics by allowing states and private actors to centralize massive volumes of data and deploy automated tools of monitoring, persuasion, and social sorting.
Two dimensions are particularly important for illiberalism studies. First, the capacity of AI systems to generate synthetic content at scale (from deepfakes to personalized political messaging) creates opportunities for disinformation, manipulation of public opinion, and erosion of trust in liberal-democratic institutions. Second, the infrastructural power embedded in AI supply chains, dominated by a few private and state-backed actors, introduces new vulnerabilities for democratic oversight and accountability. These developments intensify concerns familiar in the literature on algorithmic governance but extend them to new levels of sophistication and potential impact.
[AI’s] scale, opacity, and dependence on centralized datasets create strong incentives for authoritarian uses—especially for the New Right whose various factions are keenly aware of the possibilities.
Indeed, AI can facilitate pervasive surveillance, population monitoring, and predictive policing, as seen in China’s algorithmic governance systems. But even within liberal democracies, the adoption of AI by security agencies raises questions about civil liberties and state overreach. The controversies surrounding Palantir’s predictive policing systems and their use in immigration enforcement illustrate how AI-enabled technologies can target marginalized groups and be leveraged for illiberal ends. Therefore, an essential question concerns the intrinsic or structural compatibility of AI with authoritarianism and illiberalism. While the potentially authoritarian nature of the technology itself is still discussed in the academic literature, its scale, opacity, and dependence on centralized datasets create strong incentives for authoritarian uses—especially for the New Right whose various factions are keenly aware of the possibilities offered by AI.
AI and the New Right
Artificial intelligence has rapidly become an object of strategic and ideological tension within the New Right, simultaneously viewed as a threat to conservative values or the embodiment of a new frontier to be explored for American techno-nationalism, but above all, an extremely powerful tool of indoctrination that is a major stake in the cultural battle against the left. To understand this ambivalence, it is useful to situate AI within the intellectual horizon outlined by Adrian Vermeule in Integration from Within (2018). Building on his account of state power, Vermeule argues that the vast bureaucratic apparatus produced by liberalism—originally designed to pursue a mirage of depoliticized governance—is not something to be destroyed but actually a tool provided “by the invisible hand of Providence” to promote new ends like the revival of common good. His strategic proposal, inspired by the figure of Saint Cecilia, is explicitly pragmatic and adaptive: illiberal actors should work from within existing institutions, infusing them with substantive Christian and postliberal values. If applied to the politics of AI, this strategy suggests that New Right actors may seek not only to capture the state’s regulatory capacities to “direct technology toward the flourishing of the family and the human person” but also to shape the ideological and normative horizons of major technology firms, embedding illiberal ideology within the design, governance, and deployment of AI systems.
To grasp the multiple intersections between AI and illiberalism, the conceptual framework proposed by Marlène Laruelle offers an analytical guide. Laruelle identifies three core “scripts” or pillars of liberalism—political liberalism, societal/cultural liberalism, and economic liberalism—that are systematically contested by illiberal actors. AI interacts with and potentially destabilizes each of these pillars in distinctive ways, making it a privileged arena for examining how the New Right and broader illiberal movements seek to appropriate technological infrastructures for ideological and political ends.
First, regarding political liberalism, AI raises concerns as its deployment has the potential to unsettle core liberal principles (such as individual rights, equality before the law, and constitutional constraint) while fostering the perception that algorithmic systems operate above the law, beyond the reach of democratic oversight or judicial scrutiny. Debates over anonymity, copyright, and liability frameworks reveal broader tensions between tech-companies and liberal regulatory norms. Calls to dismantle fact-checking infrastructures, weaken antitrust enforcement, or circumvent parliamentary controls reflect an illiberal political culture that positions technological innovation as threatened by democratic accountability.
In that vein, at the 2025 Munich Security Conference, J. D. Vance explicitly framed regulation of AI as a threat to freedom and innovation, arguing that “excessive regulation” would kill the transformative potential of AI and warning European democracies against imposing what he called undue constraints on speech and technology. He used this platform to decry what he perceived as creeping censorship across Europe, asserting that new rules on digital speech and data governance represented a rollback of fundamental liberties under the pretext of security. Vance’s rhetoric implicitly challenges democratic and parliamentary oversight over data and AI: by insisting on deregulation, he signals a preference for technological actors having free rein over data and algorithmic systems, potentially bypassing the oversight role of parliaments and democratic institutions.
By insisting on deregulation, [J. D. Vance] signals a preference for technological actors having free rein over data and algorithmic systems, potentially bypassing the oversight role of parliaments and democratic institutions.
Second, concerning societal liberalism, AI platforms have become battlegrounds for contesting pluralism, multiculturalism, and inclusion. Changes in moderation norms on large-scale models (exemplified by the loosening of safety constraints in certain AI systems) have amplified the visibility of hateful, exclusionary, or anti-diversity narratives. These dynamics align with New Right discourses that oppose the core tenets of cultural liberalism and view “ideological neutrality” in technology as a justification for reintroducing discriminatory or exclusionary speech into mainstream digital spaces.
Third, in relation to economic liberalism, the political economy of AI raises profound questions. The sector is dominated by a small number of firms, leading scholars to describe it as a form of “monopoly capitalism” or “crony capitalism” driven by close ties between political elites and tech companies. While many AI firms advocate deregulation in the name of innovation, they simultaneously rely heavily on public subsidies, state-provided data infrastructures, and preferential regulatory environments. This combination (state support without state oversight) aligns with a neoliberal yet illiberal agenda. The Trump administration’s pledges of massive AI investment paired with deregulation exemplify this configuration, revealing how the New Right may see AI both as a strategic economic tool and as a mean to weaken liberal democracy.
Conclusion
AI (and technological innovations more globally) must now be treated as a central component of contemporary debates on illiberalism. It occupies an increasingly prominent place in the political agenda of the U.S. New Right and is beginning to structure similar conversations in Europe. AI is also gaining traction within illiberal thought collective, where it is discussed not only as a tool of cultural transformation but as a strategic lever for reshaping institutional and economic orders. For these reasons, the study of illiberalism can no longer bracket out AI. Understanding how the New Right imagine, instrumentalize, and seek to govern AI is essential for grasping the evolving landscape of illiberal ideology.
Raphaël Demias-Morisset (Ph.D, Political science) is a Research Fellow at the University of Bordeaux and a Research Associate at the Illiberalism Studies Program, where he works on the Technology and illiberalism Project. His main fields of research are the history of liberal and conservative ideas, constitutional theory and the study of ideologies, and his research focuses mainly on the conceptualization of illiberalism and its relation with neoliberalism, and illiberal theories. He published a book in France in October 2025 titled “Illiberalism: The Ideology of the New Conservative Revolution.”
Image “Mobile phone with ChatGPT on keyboard (52916924616)” by Jernej Furman licensed under CC BY 2.0.
