1. Governing Artificial Intelligence?

Artificial intelligence (AI) has become a focal point for contemporary debates about power, normativity, and the shaping of collective futures. Once confined to expert domains of computer science and engineering (e.g., Dreyfus 1972; Barr and Feigenbaum 1981), AI now permeates everyday life: from recommender systems and biometric identification to generative language models, creative tools, and automated decision-making in governance, labor, and culture (e.g., welfare fraud detection systems, automated résumé screening and ranking, or algorithmic management platforms). This expansion has been accompanied by an equally rapid proliferation of normative claims. From a techno-optimist standpoint, AI promises efficiency, innovation, and social progress (Chow and Celis Bueno 2025), while at the same time raising concerns about surveillance, discrimination, labor displacement, epistemic authority, and democratic legitimacy (Zuboff 2022; Taeihagh 2025). Against this backdrop, AI increasingly appears not merely as a technological artifact but as a political problem that is itself productive of new forms of governance, regulation, critique, and contestation.

The guiding question of this special collection, "Governing Artificial Intelligence?," is intentionally ambivalent. On the one hand, it refers to ongoing attempts to govern and control AI through ethical guidelines, legal frameworks, institutional oversight, and technical design principles intended to embed values such as safety, accountability, and fairness at the stage of creation. These efforts take place alongside influential positions within the technology sector that promote libertarian or laissez-faire approaches, opposing regulation and prioritizing innovation over control (Andreessen 2023). On the other hand, the special collection's guiding question asks how AI itself participates in governing: how it shapes conduct, redistributes agency, structures fields of possibility, and reorganizes relations between humans, technologies, and other nonhuman actors. This dual meaning resonates with a long-standing concern in critical theory and critical genealogy—the notion that power does not merely prohibit or repress but operates productively, through dispositifs (Foucault 1994) that orient behavior, normalize expectations, and configure environments.

Rather than treating AI as a neutral tool that can be aligned with predefined human values, this special collection approaches it as a mode of shaping conduct and of structuring processes of subjectivation—a diverse ensemble of practices, imaginaries, infrastructures, and norms that actively influence how individuals understand themselves, their capacities, and their relations to others. For instance, AI-driven social media algorithms privilege forms of visibility and engagement that encourage users to cultivate themselves as recognizable, measurable, and continuously evaluable subjects. Drawing from media studies, philosophy, cultural analysis, legal scholarship, and critical social theory, the contributions assembled here interrogate how AI reconfigures ethical, cultural, and political relations. Taken together, they explore not only how AI should be governed, but also how AI governs us—and how these two dimensions might be disentangled, rethought, or transformed.

To articulate this problem space, we propose the concept of more-than-human governance as the shared theoretical framework of the special collection. Drawing on Michel Foucault's analytics of governmentality as interpreted by Thomas Lemke (2021), we argue that AI governance cannot be adequately understood within an anthropocentric framework that presupposes a stable, sovereign human subject overseeing technological processes. In this sense, we critically engage with recent calls for Digital Humanism (Werthner et al. 2022 and 2024), arguing that many such approaches—precisely through their ethical framing—risk reinscribing the very humanist assumptions that Foucault's work invites us to question.

2. Configuring More-than-Human Relations

The notion of governance has long been associated with the regulation of human populations, institutions, and behaviors. In Michel Foucault's lectures on governmentality, however, governance is redefined as a broader, more diverse set of practices concerned with guiding and shaping conduct (Foucault 2009). Apart from functioning through laws or sovereign commands, government thus operates through heterogeneous arrangements that align subjects, knowledges, material infrastructures, and environments. These arrangements steer how people behave, define what counts as knowledge or truth, structure physical and digital spaces, shape which actions are easy or difficult. Importantly, Foucault's later work gestures toward what he calls a "government of things" (Foucault 2009, 97): a form of power that targets not individuals or populations, but rather the relations between humans, nonhumans, and their milieus.

Thomas Lemke's interpretation of Foucault radicalizes this insight by developing explicitly more-than-human analytics of government (Lemke 2021). In contrast to readings that portray Foucault as confined to the social or to discursive constructions, Lemke emphasizes the Foucauldian concepts of dispositive, technology, and milieu as analytical tools for examining how governance functions through material entanglements. "My thesis is," he states, "that putting forward the notion of the dispositive, a comprehensive understanding of technology, and a complex reading of the milieu provides elements for a thoroughly relational materialism" (Lemke 2021, 12). From this perspective, governance is less about imposing norms on preexisting subjects than about arranging heterogeneous entities—such as bodies, algorithms, data infrastructures, legal standards, affects—so that certain forms of conduct become more likely than others.

AI is exemplary of such arrangements. Machine-learning systems do not simply execute commands; they modulate environments, pre-structure decisions, and recalibrate feedback loops between users, institutions, and infrastructures. Recommendation algorithms reorder attention and desire; biometric systems reconfigure regimes of visibility; generative models reshape communicative practices and epistemic authority. In each case, AI participates in governing by shaping the conditions under which actions, choices, and judgments take place—in other words, it forms the environment of decision-making. Governance thus occurs through AI rather than merely over it.

Understanding AI in these terms requires abandoning a narrow focus on intentional human control and embracing a relational view of agency. Agency, in a more-than-human sense, is distributed across networks of humans and nonhumans; it emerges from interactions rather than residing in a single actor (Latour 2005). This does not entail denying human responsibility or political accountability. Rather, it invites a shift in focus from questions of mastery ("Who controls AI?") to questions of configuration ("How are human and nonhuman capacities aligned, constrained, or amplified?"). By aligning Foucauldian governmentality with insights from science and technology studies (STS), Lemke proposes a "more-than-human analytics of government" that "acknowledges and attends to the complexities of material entanglements while still endorsing the concept of a strong responsibility that accounts for the vulnerabilities, injustices, and hazards humans inflict on both humans and nonhumans" (Lemke 2021, 157).

Crucially, this perspective also reframes ethical concerns. Instead of asking how to encode abstract values into autonomous systems, a more-than-human approach investigates how ethical norms are enacted, transformed, or undermined in concrete practices. Ethics becomes less a matter of principles and more a matter of arrangements: of how specific sociotechnical configurations enable or foreclose certain ways of living, relating, and acting.

3. Questioning Digital Humanism

Recent debates on AI governance have been strongly shaped by calls for Digital Humanism. Promoted prominently in European policy and academic contexts, this approach seeks to ensure that digital technologies—and AI in particular—serve fundamental rights and democratic ideals (Werthner et al. 2024). As articulated in initiatives such as the Vienna Manifesto (Werthner et al. 2022, xi–xii), Digital Humanism advocates the development of technologies aligned with these principles, emphasizing human oversight, accountability, and the preservation of human agency in increasingly automated environments. Insisting on human responsibility and dignity, Digital Humanism positions itself as an alternative to technological determinism and market-driven innovation (Díaz de la Cruz et al. 2025).

From a genealogical perspective, several interrelated limitations can be identified in this framework. First of all, as Foucault repeatedly argued, humanism is not a universal or self-evident foundation but a historically specific discourse that produces particular notions of "the human" and its supposed essence (Foucault 1971). By taking "the human" as a normative anchor, humanist frameworks risk obscuring the contingency and heterogeneity of subjectivity, as well as the power relations through which certain forms of humanity (e.g., white, male, Western) are privileged over others (Lucci and Osti 2024). Such asymmetries can become embedded in sociotechnical configurations, for instance, in datasets, design practices, or institutional norms. Instead of being neutral, automated hiring or productivity systems may privilege styles of communication and performance historically associated with dominant social groups.

Digital Humanism tends to reproduce this assumption by presupposing a stable human subject that must be protected from technological encroachment. "We must shape technologies in accordance with human values and needs, instead of allowing technologies to shape humans," states the Vienna Manifesto (Werthner et al. 2022, xii). From this perspective, AI appears as an external force that threatens to displace or diminish human agency, creativity, or judgment—unless it is carefully regulated and aligned with human values. In a more-than-human view, however, this framing is misleading. Humans and technologies have never been ontologically separate; they have always coevolved through practices, tools, and institutions. The question, therefore, is not how to keep AI "in its place" but how different configurations of human–machine relations produce different modes of experience, subjectivation, and responsibility.

Lemke's reading of Foucault is instructive here, because it shows how appeals to the human can function as techniques of governance rather than as its antidote. By defining what counts as properly human—rational, autonomous, responsible—humanist discourses can marginalize those who do not fit this model, while simultaneously legitimizing specific forms of intervention and control. The concept of "the human" then acts as a filter that determines who is considered fully legitimate and who is seen as deficient. In the context of AI, Digital Humanism thus narrows the field of critique by focusing on safeguarding a particular notion of the human instead of interrogating the broader assemblages through which AI participates in configuring more-than-human assemblages, including plants, animals, minerals, etc. (Buongiorno and Chiaramonte 2024).

Hence, this anthropocentric orientation has the effect of marginalizing ecological relations within dominant frameworks of AI governance. By grounding critique primarily in the protection of human autonomy, dignity, and agency, approaches such as Digital Humanism tend to background the material and environmental conditions through which AI systems are produced and sustained. As critical work on the planetary costs of AI has shown, contemporary AI infrastructures are inseparable from extractive economies, energy-intensive computation, and global supply chains of minerals and labor (Crawford 2021). From a genealogical perspective, the relative absence of ecological sustainability in human-centered discourses is therefore not accidental but symptomatic: environmental impacts are frequently treated as secondary concerns rather than as constitutive elements of governance (Beck et al. 2021). A more-than-human approach challenges this separation by insisting that ecological relations are integral to the configurations of power, responsibility, and conduct through which AI operates.

4. Exploring Sites of AI Governance

The five contributions to this special collection do not approach the governance of AI as a single institutional problem but as a constellation of situated practices in which norms, values, affects, and power relations are continuously enacted. Rather than treating governance as the exclusive domain of policymaking or regulation, the articles focus on diverse cultural, ethical, legal, and institutional sites in which AI becomes governable—and governing—through everyday practices, imaginaries, and infrastructures. These settings range from popular television and academic labor to regulatory frameworks, creative industries, and ethical theory, revealing such governance as a dispersed, multi-layered, and often internally contradictory process. In this sense, AI governance can also be understood as operating through sociotechnical imaginaries that structure collective expectations of technological futures (Jasanoff and Kim 2015).

Jaimey Fisher's essay (2024), "'I won't bite': Generative AI, Robotics, and the Ethics of Loss in Black Mirror 201 'Be Right Back,'" situates AI governance within the realm of popular culture, where moral intuitions are rehearsed and contested before formal regulation takes shape. By analyzing the "Be Right Back" episode of the TV series Black Mirror, Fisher shows how generative AI and humanoid robotics become affective technologies that reorganize experiences of grief, intimacy, and memory. Through narrative strategies of focalization and alignment, the episode draws viewers into ethically ambiguous situations without resolving them into stable judgments. The AI-generated replica of a deceased partner does not simply raise abstract questions about consent or authenticity; it actively governs emotional conduct by offering consolation, provoking discomfort, and redefining what counts as acceptable responses to loss. In this sense, popular media emerges as a crucial site of AI governance: not because it prescribes norms, but because it shapes moral sensibilities, emotional repertoires, and shared imaginaries through which human–machine relations are evaluated.

Michael Litschka's article (2025), "AI Ethics and the Capability Approach," shifts the focus from cultural imagination to normative theory, yet arrives at a similarly relational understanding of governance. Drawing on Amartya Sen's capability approach, the article critiques dominant ethical frameworks that reduce AI governance to either utility maximization or abstract principles such as transparency or fairness. Instead, it proposes evaluating AI systems based on whether they expand or constrain the real freedoms of individuals and collectives. Litschka emphasizes that capabilities are not inherent qualities of isolated subjects but emerge from institutional, educational, and technological arrangements. From this perspective, AI governance is inseparable from questions of media literacy, organizational responsibility, and public reason. The article thus reorients AI ethics away from the attribution of moral properties to machines and toward the sociotechnical conditions under which agency, participation, and justice become possible.

In "Ethics in Regulating Artificial Intelligence: An Overview of the Recent Legislation in the European Union," Krisztina Rozgonyi, Mari-Liisa Parder, and Rodrigo Conde Jiménez (2026) analyze contemporary European AI regulation as a key site of ethical governance. Focusing on instruments such as the EU AI Act and related digital governance frameworks, the article traces how principles of transparency, accountability, and non-discrimination are operationalized through risk categories, technical standards, and compliance mechanisms, while other values—such as democracy, solidarity, or sustainability—remain weakly institutionalized or are relegated to non-binding commitments. This analysis demonstrates that governance is a process of prioritization and omission: ethical concepts do not merely guide regulation but are actively reshaped, narrowed, and reordered in their passage into law and technical design. AI governance appears, then, not as the neutral application of shared values, but rather as a field of struggle over which concerns count as governable and which remain marginal.

The contribution by Yulia Belinskaya, Christian Holst, and Olga Kolokytha (2026), "From Muse to Machine: A Scoping Literature Review of AI in Cultural Production," examines AI governance in the scholarly and practical fields of creative industries. Rather than framing AI as a threat to human originality or authorship, the article conceptualizes creativity as a distributed practice involving artists, datasets, algorithms, platforms, and legal regimes. Questions of authorship, ownership, and aesthetic value are shown to be renegotiated within these assemblages, often in ways that reconfigure power relations between creators, technology providers, and cultural institutions. Governance here operates not primarily through prohibition, but through infrastructural and economic arrangements that shape access, visibility, and attribution in more-than-human creative ecologies.

Finally, Joan Ramon Rodriguez-Amat's essay (2026), "Pleasure, Happiness, and Diversion: Doing a PhD in Times of Artificial Intelligence," turns the analytical lens toward academia as a site of AI governance. Through an essayistic and autoethnographic mode of writing, the article explores how generative AI intersects with metric-driven evaluation systems, bureaucratic rationalization, and neoliberal regimes of productivity. The author argues that AI governance in academia increasingly operates through optimization logics that privilege efficiency and output over intellectual risk, curiosity, and collective engagement. Against this backdrop, he defends the PhD as not merely a training in skills, but as a space for pleasure, happiness, and diversion—values that resist algorithmic standardization. The essay thus considers academic labor to be a contested terrain where AI governs conduct by shaping which types of thinking are encouraged, rewarded, or rendered obsolete.

Taken together, these five contributions demonstrate that AI governance unfolds across different settings in which human and nonhuman actors are entangled in complex ways. Governance is neither centralized nor monolithic; it is enacted through narratives, norms, infrastructures, and affects that configure possibilities for action and judgment. Hence, the special collection invites readers to consider AI governance not only where it is explicitly named, but also where it is silently embedded in cultural practices, institutional routines, and everyday forms of conduct.

5. Governing AI Otherwise?

Building on the theoretical framework and empirical analyses presented above, the question arises of how AI governance might be practiced otherwise. Governing-AI-otherwise does not mean rejecting regulation, ethics, or oversight; rather, it entails rethinking their premises and limits. From a Foucauldian perspective, this involves shifting the focus from the search for universal principles or definitive solutions to the analysis of concrete regimes through which power operates. AI governance becomes an ongoing, contested practice of arranging relations between humans and machines, but also between values, institutions, and diverse forms of life.

A more-than-human approach emphasizes the fact that AI does not enter a neutral social space but actively reshapes the world it operates in, affecting the milieus in which action and judgment occur. Governing-AI-otherwise therefore requires sensitivity to how these milieus are produced and modified: through technical standards and interfaces, through cultural representations and affective economies, through educational systems and labor regimes. It also demands attentiveness to asymmetries of power that persist within supposedly decentralized or automated systems, as evident in the ownership of infrastructures, the definition of algorithms and standards, or epistemic authorship. Distributed agency does not imply distributed responsibility: when diverse actors and automated processes are involved, responsibility becomes blurred, calling for new forms of accountability that are attuned to relational and infrastructural dynamics. Rather than aiming to prevent a certain outcome, we should ask which settings of algorithms, institutions, and design choices produce what effects, rendering power relations and decision-making traceable and assignable at different control points.

In this vein, the critique of Digital Humanism stated above assumes particular importance. Appeals to "human-centered AI" or the primacy of human agency can serve as rhetorical shortcuts that mask the processes through which subjectivities are shaped and governed. Rather than anchoring governance in an abstract notion of humanism, governing-AI-otherwise entails treating "the human" itself as a variable—produced, destabilized, and transformed within sociotechnical assemblages. This does not negate ethical concerns for dignity or justice; it situates such concerns within a historical and political analysis of how they are mobilized, operationalized, and, at times, instrumentalized.

The contributions to this special collection do not converge on a single normative program, nor do they prescribe a unified model of AI governance. Their commonality lies instead in a shared methodological orientation: a commitment to analyzing AI governance as a set of practices that shape conduct, redistribute capacities, and redefine relations. This genealogical approach resists both technological determinism and moralism. It acknowledges that AI is neither an inevitable destiny nor a mere tool, but a contingent and contested dispositive whose effects depend on the configuration of its social, cultural, and political relations.

At the same time, this perspective raises an important concern: if governance is understood as encompassing cultural imaginaries, technical infrastructures, emotional dispositions, educational practices, legal regulations, and diverse forms of life, does the framework risk becoming too expansive for practical use—particularly in policy or operational contexts that demand clarity, prioritization, and timely decision-making? A more-than-human approach, however, need not function as an all-encompassing or totalizing description. Rather, it can be understood as a diagnostic mode of analysis. Such an approach does not assume that all elements are equally relevant in every situation; instead, it examines how specific configurations of heterogeneous actors and relations coalesce in particular settings to produce concrete effects. By tracing how agency, responsibility, and power are distributed within these assemblages, it becomes possible to identify points of leverage, intervention, and accountability. In this sense, a more-than-human perspective enhances rather than diminishes analytical precision, helping to avoid oversimplification, "black-box" reasoning, and misplaced responsibility by redirecting attention from immediately visible surface phenomena to the underlying relational and infrastructural pressure points through which governance operates.

Governing-AI-otherwise, then, is less about mastering technology than about cultivating critical reflexivity toward the conditions under which AI operates. It involves opening spaces for experimentation, dissent, and alternative imaginaries—whether in media representations, academic practices, regulatory debates, or creative production. By advancing the notion of more-than-human relations, this special collection seeks to contribute to such a reflexive practice of governance: one that remains attentive to its own limits and open to the possibility that AI governance, like AI itself, might always be otherwise.

AI Declaration Statement

Generative AI technologies, namely Microsoft Copilot and DeepL, were used selectively for language editing and drafting support during the preparation of this article. All conceptual arguments, interpretations, and revisions were developed and critically assessed by the authors.

Competing Interests

The authors have no competing interests to declare.

Authors' Contributions

Simon Ganahl initiated and coordinated the special collection "Governing Artificial Intelligence?," oversaw the editorial process of the included articles, and drafted the initial version of this introduction. Yulia Belinskaya substantially revised and refined the introduction and contributed equally to the editorial work on the collection. Simon Ganahl serves as the corresponding author.