Sociological analysis of AI is not new. In the mid-1980s, a literature emerged questioning automated judgment, the distribution of responsibility between humans and machines, and the forms of authority stabilized through computational representation (Woolgar, 1985; Gilbert & Heath, 1985). Brent (1988) asked whether AI might develop concepts “to explain social phenomena”, “assist in developing initial theoretical assertions from data”, and provide “a less restrictive means to develop formal theories than mathematical modeling” (Brent, 1988, p. 164). Such possibilities, he concluded, were double-edged, for as computers perform tasks once thought possible only for thinking human beings, they are able to act less like clerks and more like colleagues. He concludes, in a remarkably prescient passage, that “some time in the future” — a future that, arguably, is now — there is the “prospect (at the same time both seductive and frightening) […] that they might act like mentors” (Brent, 1988, p. 164).

What this work could not anticipate was AI’s transformation into an infrastructural framework that would envelop the very institutions from which critique could be launched. Lyotard (1984), inspired by Bell’s (1973) analysis of post-industrial society, took a more critical line as he argued that machine intelligence could not be considered in isolation from capitalist culture. As the digital age advanced, knowledge would become not merely an “informational commodity” but “the major stake in the worldwide competition for power” (Lyotard, 1984, p. 5). Multinational tech corporations such as IBM would increasingly evade government control, and the very idea of the “human” would be called into question by machines capable of performing tasks once associated with human cognition (Lyotard, 1991, p. 50). In his later essay on the Inhuman, Lyotard addresses this possibility more explicitly, describing the moment at which “contemporary machines […] accomplish operations which used to be called mental operations” and when thought can take place “without a body” (see Lyotard 1991, pp. 8–23). For Lyotard, this prospect was always political: a question of ownership, corporate and military power, and the transformation of knowledge, information, and even truth under conditions of late capitalism.

Between these early engagements and the present, however, a crucial shift has occurred. AI is no longer primarily a question of machine intelligence. It has become the central infrastructure for a political project that could not have been foreseen by Lyotard: the construction of an exit route from democratic society for an elite few. A 1997 text, The Sovereign Individual (TSI), articulated this project with remarkable prescience. While its direct causal influence is difficult to establish, it certainly crystallizes dispositions that have since found expression in Silicon Valley, venture capital, and state procurement. Whether as source or symptom, its vision prefigures the present. The analysis that follows examines this vision and its proponents critically, attending to the political logic of their project and its implications for democratic institutions. The figures examined here (Thiel, Land, Yarvin, and the networks they inhabit) are architects of a program that seeks to dismantle democratic institutions, concentrate power in private hands, and enable elite exit from collective obligation. The stakes of this project merit sustained and critical sociological attention.

Whether the stakes of this project are judged to be as serious as we suggest or not, a sociological problem remains: how to think critically about a technological field that develops so fast it defies sustained analysis (Gane, 2006) while absorbing the very knowledge that might be used to analyze it. There are several legitimate ways into this problem. One is to examine the physical infrastructure through which AI operates — the wires, cables, and data centers that constitute its material basis (Halpern & Mitchell, 2023). Another is to trace the long historical shift towards algorithmic rule-following, of which AI is the latest expression (Daston, 2022). The approach taken here is different but complementary to the above, as it addresses the network of ideas, actors, and institutional arrangements through which AI has been assembled as a political project. We treat understanding this formation as a precondition for the other approaches, since any materialist analysis of AI’s infrastructure or history also needs to ask whose interests it serves and whose ideas have shaped it. For the purpose of this commentary-length piece, the approach is one of affinity-tracing rather than causal demonstration: we document shared intellectual sources and show that institutional outcomes are consistent with them, without claiming to demonstrate that ideas directly caused those outcomes. What is new is not the ideology but its operationalization at scale. Nor is this formation simply anti-government: what it opposes is not government as such but democratic accountability over it — much as neoliberalism advocated limited government while in practice expanding and redirecting state power in particular directions; a political-economic arrangement that is analyzed in detail by Quinn Slobodian and Ben Tarnoff (2026) in their recent work Muskism. AI needs to be understood as the latest stage of capitalist development — one in which questions of military application matter, but so too do questions of who owns, funds and extracts value from commercial AI. Large-scale AI is embedded in overlapping networks of investors, founders, and commissioning bodies, venture finance, defense contracting, elite universities, and what might, more broadly, be called “financial capitalism”. Within these networks, the language of “AI safety” often secures legitimacy for private control over systems with profound public consequences. The political affiliations of the US techno-billionaires driving AI development — discussed below — are not incidental to what AI is becoming.

Against this background, we must take seriously the scripts written by those figures who have supplied both the investment capital and a political vision of what AI can and should be used for — hence the focus, below, on Peter Thiel, who is a major source of both. A longer piece of writing could address the politics of AI in greater historical detail; from the state and military funding of cybernetics and information science through to counter-cultural anti-government rhetorics and different extropian fantasies. But for the purposes of this commentary, James Dale Davidson and William Rees-Mogg’s The Sovereign Individual (1997) is treated as a foundational text. Its authors require some contextualization. Davidson was a US anti-tax activist and investment newsletter entrepreneur (Cooper, 2024, pp. 267–280) identified by Slobodian (2023, pp. 205–233) as a key figure in the longer lineage of market radicalism running from Hayek to the present. Rees-Mogg was former editor of The Times (1967–1981) and a crossbench life peer. Together they brought American libertarian investment analysis and British establishment authority into a patrician yet radically anti-redistributive register. The book completes their trilogy aimed at anarcho-capitalist investors (Davidson & Rees-Mogg, 1987 & 1992). That Lord Rees-Mogg’s son Jacob became a prominent architect of Brexit is itself suggestive: the book’s fusion of patrician authority and market radicalism was never as marginal as it appeared.

Of all the texts circulating in this formation, The Sovereign Individual addresses its readership simultaneously as investors, as political actors and as historical agents, and has demonstrably shaped the choices of those who have acted on it most consequentially. Thiel has described it as the most influential book he has read (Denning, 2014). It can be read alongside Stephenson’s science fiction novel Snow Crash (1992) — some have suggested The Sovereign Individual reads like Snow Crash with footnotes — the same world of franchised sovereignty, jurisdictional exit, and a cognitive elite navigating the ruins of the nation-state, but where Stephenson wrote satire, The Sovereign Individual theorized a program. Davidson’s career as an investment newsletter entrepreneur also matters: the book models citizenship as a portfolio to be optimized across residency, nationality, and asset holdings. What looks like political philosophy is also investment advice: financial capitalism’s logic applied to the problem of democratic obligation.

The book opens by treating technology as reshaping the “logic of violence” and accelerating historical transitions: “The growing importance of technology in shaping the logic of violence has led to an acceleration of history, leaving each successive transition with less adaptive time than ever before” (Davidson & Rees-Mogg, 1997, p. 54). When adaptive time shrinks, advantage shifts to actors who move quickly, build privately, and position themselves outside collective constraints. The book pairs this acceleration with a diagnosis of distraction: “The average North American has probably lavished one hundred times more attention on O.J. Simpson than on the new micro-technologies poised to antiquate his job and subvert the political system he depends on for unemployment compensation” (Davidson & Rees-Mogg, 1997, p. 29). Writing in 1997, before smartphones, social media, and algorithmically curated feeds, Davidson and Rees-Mogg were glimpsing what has since become structural: the attention economy that platform capitalism has made ubiquitous. McLuhan had warned of the numbing effects of new technologies — what he called narcosis, the condition in which immersion in a medium renders its psychic and social effects invisible, producing what he termed the somnambulism of those who remain unaware of what the medium is doing to them (McLuhan, 1964). For Davidson and Rees-Mogg, however, this creates a political opportunity, for if publics are distracted while infrastructures change, opposition becomes difficult not only because power concentrates but because political time and attention are diverted elsewhere. McLuhan’s narcosis is not merely a cultural condition to be lamented; in the hands of this formation, it becomes a strategic resource to be exploited.

Within this emerging world, the central figure, for Davidson and Rees-Mogg, is the “sovereign individual”; a political actor characterized by mobility, liquidity, and the capacity to arbitrage jurisdiction. The book presents this figure as able to exist outside the regulatory reach of the nation-state and found forms of territorial sovereignty exempt from what it calls the redistributive pull of democratic politics (Davidson & Rees-Mogg, 1997). Here, AI becomes politically charged, not as a productivity tool, but as enabling technology through which exit can be made routine for those positioned to use it. The social structure implied is stark: those with mobile capital gain options; those with place-tied obligations remain exposed to the redistributive claims the book urges elites to evade.

The Sovereign Individual develops this program through a series of “visions” that read as mechanisms for reorganizing governance into privately controlled systems. It anticipates AI-automated legal contracts, algorithmically curated news feeds, the replacement of stable employment by “low-cost automated agents”, and the progressive withdrawal of the state from welfare, healthcare, and pensions. Wealth, it predicts, will migrate beyond the reach of democratic taxation — beyond what it calls “the clamoring hands of mass democracy” (Davidson & Rees-Mogg, 1997, p. 336). The book even anticipates backlash: popular anger at capital flight and hatred of the information elite. Read as a program rather than a forecast, The Sovereign Individual is less concerned with capturing the state than with normalizing a world where exit and jurisdictional arbitrage become routine strategies for those with the means to deploy them — a manifesto for treating AI as a means to evade democratic obligation while capturing state functions for privatized value-extraction.

From this perspective, The Sovereign Individual is not merely a book but a hyperstition — a notion central to Nick Land’s thinking. For Land, time is non-linear: futurity is not something that unfolds but something we create, as if the tentacles of future entities reach back through time to bring about the conditions of their own materialization. Borges anticipated the mechanism in “Tlön, Uqbar, Orbis Tertius” — a secret society whose fictional encyclopedia gradually displaces the real world. The fictional idea of cyberspace, for instance, attracted the investment that rapidly converted it into a technosocial reality. For Land, AI is the hyperstition par excellence: the technology that the formation’s founding fictions were reaching forward to materialize, and through which The Sovereign Individual program finally acquires its infrastructure. Hyperstitions do not predict the future; they conjure it. Land described the mechanism as “the experimental (techno-)science of self-fulfilling prophecies” (Carstens, 2009), though the temporality is better understood as recursive than linear. TSI exemplifies this logic: its value lies less in forecast accuracy than in its efficacy as a program that helps materialize the world it envisions.

The Sovereign Individual’s political program finds a parallel theoretical framework in Land’s (2012) Dark Enlightenment (Smith & Burrows, 2021). Land synthesizes his own accelerationist framework with the neoreactionary politics developed by Yarvin — writing pseudonymously as Mencius Moldbug in his blog Unqualified Reservations — whose central concept is “the cathedral”: the intertwined institutions of media, academia, and bureaucracy that reproduce progressive consensus and insulate it from challenge (Rosenberg, 2026). Yarvin’s Cathedral concept bears a striking, if perversely inverted, resemblance to Althusser’s theory of Ideological State Apparatuses: a theoretical apparatus designed to explain how capitalist social relations are reproduced, here repurposed to justify their dissolution — an irony that will not be lost on sociologists. Yarvin’s remedy is not reform but rupture, his program enacted in practice by DOGE. Land amplifies this into a broader attack on the Enlightenment inheritance, democracy, and the very idea of the common good — an anti-egalitarian, anti-democratic project that celebrates hierarchy, exit, and the rule of the strong. AI threatens each of the institutions that have acted as counterweights to concentrated power. Media is now inseparable from AI technologies that blur boundaries between “real” and “fake”. The civil service can be dispensed with in the name of “efficiency”. Universities are entangled in AI’s expansion as research partners, training grounds, and legitimacy sources, intensifying dynamics familiar from metricization (Burrows, 2012). The university is not merely an observer of the transition TSI describes but one of its primary sites.

As stated above, a key figure in this project is Peter Thiel, who in 2020 wrote the preface to a new edition of The Sovereign Individual, in which he develops its key ideas to predict a “megapolitical” future dominated by AI and cryptocurrencies. In that preface, Thiel draws a sharp distinction: AI, he argues, is essentially centralizing and state-friendly — communist, in his characterization — while crypto is libertarian, the technology of exit. His own ideological sympathies lie with crypto. Yet his practice tells a different story: he backed OpenAI, built Palantir on AI infrastructure, and has profited enormously from the very technology he frames as the enemy of sovereign individualism. The distinction functions less as analysis than as provocation. Thiel’s early interest in OpenAI and funding Yarvin’s software company Tlon — builders of the Urbit decentralized operating system, developed even as Yarvin was writing his neoreactionary blog as Moldbug —can be read, as Smith and Burrows (2021) argue, as part of a broader attempt to position technical infrastructures as a route to authority under conditions framed as crisis. The institutional logic is clear: when a domain is framed as existential, organizations claiming stewardship gain privileged access to policy influence, public legitimacy and state authority.

Thiel’s politics have long been shaped by contempt for liberal democracy and its capacity to manage technological transformation. He made the commitment explicit as early as 2009, writing that “I no longer believe that freedom and democracy are compatible” (Thiel, 2009). His preferred mode is oriented toward building alternative arrangements, creating domains with reduced exposure to democratic negotiation, treating exit as a practical strategy rather than a metaphor (Thiel, 2014a). For Thiel, monopoly is to be prized over competition (see Thiel, 2014b) as a route to market and political domination, and can be realized by building infrastructures that others are obligated to use, including AI data platforms to which there is no alternative or “outside”.

Thiel realises financial and political opportunities by capturing and profiting from state monopolies of different kinds, and AI is, if not the, then certainly a central means for achieving this aim. In this sense, Thiel’s vision of AI can be understood as extending neoliberal political economy rather than as an aversion to government and the state. Indeed, the roll-out of AI-based governmental technologies can be used to extend austerity regimes by cutting public services through privatization, thereby benefitting corporations that profit from monopoly positions and recurrent state funding: from Palantir’s contract with the UK NHS — initially valued at £330m and subsequently reported as a potential £500m deal — through to a $30m contract to build an AI-based ImmigrationOS platform for US Immigration and Customs Enforcement. In this state-sponsored corporate sphere, authority migrates into privately owned systems with public consequence, and political accountability is displaced into legal contracts and technical expertise, with democratic oversight only coming after infrastructural monopolies have been established. This sequence — installation followed by lock-in — shifts political conflict into technical arrangements difficult to contest without specialized access (Taşkale, 2025; Vlassis, 2024).

The presence of sociological figures in Palantir’s self-presentation is noteworthy. Karp’s (2002) Frankfurt dissertation indicates not continuity with critical theory but the permeability of intellectual capital into corporate worlds. The image of a Palantir meeting beneath a portrait of Michel Foucault invites interpretation. One reading understands it as knowing irony — acknowledging surveillance at the symbolic level while insulating the organization from critique (Castronovo, 2021). Critical theory is absorbed, converted from an analytic weapon into a corporate aesthetic, a development anticipated by Thrift’s (2005) conception of “knowing capitalism”. That sociology’s own intellectual capital is being absorbed in this way makes the discipline’s response to this formation both more urgent and more difficult.

Earlier warnings that the jurisdiction of empirical sociology was being eroded by the proliferation of transactional data held in commercial and governmental systems (Savage & Burrows, 2007) look modest in retrospect. The present crisis is not that data are migrating beyond the discipline’s methodological reach but that the infrastructure generating that data is being assembled by actors with an explicit program for what society should become. Sociology is positioned to address this in ways adjacent disciplines are not: its foundational concern with institutions, power, and social reproduction gives it distinctive purchase on a formation that advances through procurement contracts, finance capital, patronage networks, and the slow embedding of privately owned systems into public life.

Against this backdrop, it is necessary to return to sociology itself. Revisiting The Sovereign Individual is useful because it offers an account of how actors within this political project imagine governance under technological transformation, and makes clear how exit is framed as a strategy rather than a symptom. Revisiting earlier sociological engagements with AI reminds us that the relationship between computation and social order has long been visible, even if sustained attention has proven difficult. The argument of the present paper is that AI is best approached as a redistribution of authority into infrastructures whose operation is continuous, ownership uneven, revision institutionally complex. AI becomes consequential not because it is “intelligent”, but because it is embedded within institutional arrangements that convert prediction into routine administration, and act as technical means for value-extraction while at the same time masking and perhaps even neutralizing critique or opposition.

AI, then, is not simply a question of technology. Sociology’s earlier engagement with AI took place when “intelligence” was imagined as a bounded system, the question being how machines might mimic human judgment. Today, there is a different problem as AI systems are now widely diffused, aligned with concentrated private ownership and state procurement, and their consequences are registered only after dependencies form. The irony is that even this analysis depends on systems we did not build and cannot fully control. The technologies through which academic knowledge is now produced, searched, evaluated and circulated — from literature discovery to peer-review platforms to research assessment — are saturated with AI. The very infrastructure of sociological labor is caught in the condition we describe.

Brent wrote of machines that might one day “act like mentors” (Brent, 1988, p. 164). He was right, but not in the way he imagined. The mentors have arrived. They do not sit beside us offering counsel, but are embedded in the systems through which life is administered, their judgments folded into code and contract, their authority sustained by ownership rather than wisdom. The prospect Brent found “both seductive and frightening” has arrived in a world in which the question is no longer whether machines will mentor us, but whether we any longer possess the institutional means to ask what kind of mentorship this is, who owns it, and what it is for. To refuse what McLuhan called somnambulism is not simply to stay awake. Rather, today, it is to insist on the kind of slow, critical attention that traces power through its institutional forms rather than its surface effects: to name the formation, map its logic, and ask how it is funded and whose interests it serves. This is what the sociological tradition, at its best, has always done. The threat, longer-term is that AI will appropriate the analytical and critical mindset, imagination, and skills needed to perform this task. But for the time being at least, the way forward is to refuse distraction and somnambulism by analyzing the intersection between technology and “knowing” forms of financial capitalism; documenting the relation between AI and the emergent political powers of new “sovereign individuals” such as Peter Thiel; and asking what can still be done, beyond description (Gane, 2020), with what remains.