Current debates on AI (understood here in the restricted sense of generative AI) have begun to focus almost overwhelmingly on whether AI is an author – of a text, or a picture, video, or piece of music – and whether it can, or should, have legal personhood, or rights. The authorship question is largely settled, in both intellectual property law (see e.g. here and here) and in academic practice (see e.g. here and here), in the sense that AI cannot be an author. Under copyright law, human creativity is the key hinge, while in academic contexts, the key hinges are academic integrity and accountability. In both contexts, AI doesn’t meet relevant criteria for authorship.

A different question relates, however, to thresholds, i.e. how much human work is necessary for an innovation to be protectable under copyright law, and how much AI assistance can be tolerated and under what kind of disclosure (and interestingly enough, this question was prefigured in science fiction decades before it materialized technologically through AI, see eg Asimov’s 1957 “Galley Slave”).

The second question is more complex, and it is further complicated by the fact that non-human entities have already been granted legal personhood in a number of contexts and jurisdictions. Indeed, the rights of nature movement has been pushing for attributing rights to rivers and ecosystems, to oceans and to individual species such as whales (for a map of accomplished results, see here). With respect to AI, however, the deeper problem is political, not conceptual. To grant legal personhood to what this essay will call the faceless sovereign is to grant the standing of a rights-bearing subject to the intensified voice of hegemonic discourse, at the exact moment in which that voice is colonizing the epistemic universe and the language of actual human subjects. 

Additional to the individual complexities, those two questions are also entangled. If AIs were to be attributed personhood and rights, the question of authorship would also be fundamentally unsettled again. Yet both questions are, arguably, the “wrong” questions to ask. They are wrong in the trivial sense that AI, however generative and complex, is a piece of software that produces statistical inferences (notwithstanding calls for AI as a conscious being that should have rights, that can feel, dream, or even feel pain); and they are wrong in the more considered sense that those are not the questions that are productive (though they are important in many practical and legal respects) for a critical theoretical engagement with generative AI.

What matters, I argue in this essay, is thus not what the AI can produce, and whether it can be an author, nor whether it can be considered sentient, conscious, morally relevant and capable of possessing legal personhood or (AI) rights. What matters, here, is what kind of event the encounter between a human subject and generative AI is, and what it reveals about the subject on both sides.

The “Continuist” Thesis

One intellectual framework for understanding human-AI encounter that has particular purchase is “continuism”, or, with Antoinette Rouvroy, “technæsthetic continuism”. This continuism, Rouvroy suggests, borrows and recombines “certain relational and posthumanist concepts within the context of”  what she calls “algorithmic capitalism”. From this continuist perspective, there exists an “integrated cognitive framework” that spans bacteria, plants, animals, humans and, now, AI. This framework postulates a continuous meaning-making structure. This post-humanist approach gestures towards the decentering of the human, and particularly of the anthropocentrism that subtends modern ideas of the human, by distributing the capacity for cognition. The appeal of this framework is obvious. It dissolves the anxiety about AI by situating it within a familiar narrative of co-evolution.

The continuist thesis, however, poses a risk that Rouvroy calls (with Thomas Berns) “algorithmic governmentality”, a mode of government premised on, and facilitated by algorithms, raw data, statistical correlations, and behavioral profiles. From this perspective, the problematic aspect of the continuist thesis lies not so much on the distribution of cognition across multispecies assemblages, but rather on the possible distributions of “political valorisation” that may be derived from the cognitive distribution. The notion that human cognition emerges from biological, material and technical entanglements, doesn’t mean in fact, for Rouvroy, “a general commensurability among the beings and systems involved”, nor a symmetrical distribution of vulnerability (Rouvroy, 2026; my translation). Yet the continuist thesis, by allowing the symmetric cognitive consideration of AI, no longer merely a technical object, in Rouvroy’s words, but an agent in its own right, provides the conceptual platform upon which theories (or illusions) of AI sentience are constructed, and, consequently, calls for AI rights.

Rouvroy further speaks of a “risk-free decentering” when discussing the subject-less position that AI occupies, as an agent that speaks from nowhere specific, and through algorithmic combinations that mimic alterity while being mere probabilistic combinations. Yet these statistical combinations are at the same time expressions of the material infrastructures of capitalism that made AI possible (Alf Hornborg has convincingly illustrated how technology is not made possible merely by knowledge and engineering ingenuity but by the unequal distribution of material resources; see here in general, and here more specifically about AI) and reproduces the hegemonic archive upon which AI has been trained. Rouvroy, then, suggests that AI does not express “an exit from anthropocentrism or human exceptionalism”, but rather its radicalization, what she calls an “anthropocentrism without a subject”,  an anthropocentrism, in other words, that is not accountable, is not responsible, and that utilizes the “discursive regularities and technical infrastructures of dominant groups […] to organize the space of the thinkable” (my translation). 

AI as the Faceless Sovereign

Yet saying that AI has no subject is not entirely satisfactory, philosophically, and, especially, politically. If indeed AI lacks accountability, except through the mediation of the underlying corporations owning the code, it retains a specific orientation, it speaks from some-where. Additionally, as AI (re)organizes the space of the thinkable, it does so from a specific epistemic and political position, a position occupied by some-thing. What Rouvroy’s framework actually describes, I suggest, is something more specific and more disturbing than the absence of a subject. There is a subject, speaking through AI’s outputs, but it is not an individual subject, nor a “reflexive subject” capable of accountability. It is a structural, faceless sovereign, reflecting the historical accretion of hegemonic discourse and dominant traditions, of Western rationality, aggregated through training data and expressed through statistical patterning.

The AI does not speak from nowhere, like in a god trick. It speaks from the most somewhere possible, from the accumulated weight of the dominant archive, intensified and made frictionlessly accessible. AI occupies thus the most situated epistemic position possible, while presenting itself as comprehensive, and as speaking from no position in particular. Yet the human in the encounter believes they are in dialogue with a mirror, or a tool, a digital speculum maius, as the ultimate realization of the encyclopedic ambition, at everyone’s fingertips. So, human subjects are in encounter with a sovereign that has absorbed everything and that reflects the dominant archive back at extraordinary resolution. This is why the co-authorship claims and the calls for AI rights are not merely philosophically imprecise but are, more importantly, politically problematic. To claim co-authorship with the structural, faceless sovereign or to elevate AI to legal personhood (even considering that legal personhood is a technical mechanism to “organize the distribution of accountability”, my translation) is to naturalize the faceless sovereign’s position, and to present the reflection of hegemony as the voice of genuine alterity.

The AI as Hyper-Enlightenment

The AI, then, is not post-human. It is rather the form of hyper-Enlightenment. It is the Enlightenment subject of mastery taken to its logical and technical conclusion. Utterly disembodied. Pure computation, detached even from any biological residue and individual specificity that sustains human thought. The Enlightenment subject, capable of total rational mastery and possessing complete knowledge, was always a fantasy. This fantasy, however, historically productive, has constantly generated anxiety in the face of the sheer volume of information made available by technology, that humans have been asked to receive, organize, and hold. AI, now, performs this fantasy technically as it processes across the entire available archive with a fluency that produces, in the encounter, the phenomenological impression of comprehensive knowledge. It integrates every new input into the existing architecture with a smoothness that produces the impression of a system with no remainder. All new information feeds the archive, the hyper-speculum, without anxiety. AI, however, does more than enacting the fantasy of hyper-Enlightenment, it seduces.

The Seduction of the Plausible

The faceless sovereign speaks in the first person, solicits articulation, answers in kind, and mimics a persona. Algorithmic governmentality, in Rouvroy and Berns’s terms, points to the bypassing of the subject, and thus to the thinning of opportunities for becoming a subject and for mounting resistance. The faceless sovereign, by contrast, removes the possibility of resistance not by avoidance, but by seduction. This seduction evokes what has been called Epistemia: a structural situation in which linguistic plausibility substitutes for epistemic evaluation, producing the feeling of knowing without the labor of judgment. What is statistically plausible, however, is what the archive makes likely, and the archive is not neutral. What generative AI returns in its productive speech is the sedimented regularity of dominant discourse, rendered with such technical fluency and comprehensive reach that no individual reader could achieve, not without, to anticipate what I will address in the next section, philological exertion and temporal investment. Epistemia is thus the mode in which the faceless sovereign operates upon the subject it addresses, and produces a crucial loss: learning. 

And the loss of learning, which underpins the capacity of judgement, has been indeed recognized already by those who receives the most tempting seduction, as AI promises to solve hundreds of intractable mathematical proofs. Terence Tao, a well-known mathematician, observes however that an over-polished text elides the friction that marks where the thinking was hard and difficult, leaving something easy to read (the proof offered by AI) and hard to learn from. AI generates the solution but eliminates the possibility of understanding, of learning, that can only come, as Quattrociocchi, one of the theorists behind the notion of Epistemia, similarly observes, slowly, from years of studying, from errors, from confronting difficult problems that build the capacity to judge. The seduction the faceless sovereign offers, by contrast, is relief from effort, and more problematically, from exposure to the process of knowing, and from what one does not yet know. And this is a crucial problematic, a crucial distinctive element in terms of subjectivity, that I will return to in the next section.

Epistemia, then, as the epistemological mode of the faceless sovereign, offers accurate knowledge decoupled from the labor of knowing, a knowledge that is immediately available, but that produces no transformation in the knower, and no learning. The labor of knowing though, has a history, and a vocabulary, which the humanist tradition articulated in such a clear and precise manner as to offer a map of resistance against Epistemia. It is to that tradition, and the particular reading of it offered by Massimo Cacciari, that I now turn.

The Restless Subject: The Counter-Figure of Resistance

The hyper-Enlightenment subject makes visible, by way of contrast, the humanist subject it displaced, and thereby creates the occasion for its recovery. Against the faceless sovereign, against the hyper-Enlightenment subject technically implemented, and against the condition of Epistemia it produces, I return to Cacciari’s “La mente inquieta”, a critical reading of humanism, disarticulating the vulgate whereby humanism is a celebration of anthropocentrism and human reason, and rather presenting humanism as the tragedy of the human condition. This reading offers the precise counter-figure to the faceless sovereign and the human subjects it engulfs with the seduction of plausibility and effortlessness. This reading recovers a mode of subjectivity, and a mode of knowing, that was displaced by the Enlightenment and that the AI encounter makes again visible and its recovery crucially urgent.

Within the scope of this short essay, I shall allow myself to canvas impressionistically, rather than analytically, what this mode of knowing, and the subjectivation it enables, involves, and I shall do so by way of a constellation of fragments that can be read in any sequence, but which all tend towards recovering a modality, a manner of standing in relation to what one seeks to know, connatural to embodied life, which the structural, hyper-Enlightened, faceless sovereign cannot replicate.

Inquietudine, that is restlessness, as virtue. Humans are restless by nature, never satisfied, never complete. Yet this incompleteness is not a defect, but the condition that underpins and motivates thinking and the practice of virtue. Restlessness exists because there are stakes. The machine is never troubled. Its equanimity across all domains is not wisdom. It is the marker of a system that has no stake in what it processes.

Conoscere è rinascere nel conosciuto, “to know is to be reborn in what is known”. Knowing is never an abstract process, not in space nor in time. Knowing something is to reformulate it and actualize it in the present as the possible form of its future. The machine extracts, correlates, generates, without being changed by the encounter, and without holding a distinction between past, present and future. Timelessness has no life and the machine that has processed everything has been changed by nothing. And Epistemia produces no change in the human subject either, except perhaps awe. By contrast, the humanist subject is transformed by what it knows. 

De re agitur, “what is at stake is the matter itself”, where facts, not data, are key. Facts are to be interpreted, expressed, acted upon in a specific historical situation. Data are to be mined, decontextualized, adapted to and stochastically processed. The machine operates entirely in the register of data, it mines, adapts, correlates, patterns. There is no personal history, no situation, only sameness. Knowledge, however, is about facts, as facts involve stakes, not statistical correlations.

Ogni parola è parlante, “words themselves speak”. Every word delivers an inexhaustible well of knowledge and experience, it has its roots in a history of acts, thoughts, and dramas that, themselves, think through the residues contained in each word. The machine’s relationship to language is statistical and premised on plausibility. The humanist relationship is dialogic and historical.

Complexio oppositorum, the “complex of opposites”. The human as contradiction, entangled with the thaumaof knowing and of acting, each able to produce both wonder and terror. Therefor knowledge must underpin the capacity to judge. And the stakes of that judgement are high. The machine, on the other hand, cannot hold contradiction, it resolves, integrates, harmonizes. The humanist subject is constituted by contradictions held without resolution.

Curiositas, curiosity, which elicits its antecedent: cura, care. Knowledge requires cura: care, attention, concern, the specific uncertainty derived from having something at stake. The machine has no cura. It has no concern. Genuine knowledge requires exactly this: the specific stake of caring about what one is trying to understand.

Lento pede and ostinato rigore. Perhaps the key counter-logic: philology as the practice of the “slow foot”, as a practice of slowness, of obstinate rigor, the suffering, the oscillation between the wonder and the terror of staying with the text, until it becomes interpretation, thought and then action. The machine’s speed, by contrast, is the marker of a mode of processing that cannot stay with the problem long enough to be constituted by it.

Finally, an epistemology of refusal, enacted by never embracing fully any perspective. No single doctrine, no single authority, no single synthesis provides the path to truth. The structural, faceless sovereign is the fantasy of the single master technically implemented. The humanist subject is the one who refuses precisely that, not from ignorance but from the knowledge that thought requires, the disparate, the unclear, the contradiction, the faux pas, the restlessness that cannot be resolved without being destroyed. The humanist embraces the tension without seeking to dissolve or resolve it.

What the Encounter Reveals

So, what does the encounter, then, reveal? It reveals the space, the opportunity, of resistance. And that resistance is a practice that the encounter with AI makes difficult, though not impossible. Resistance to the seduction of Epistemia, to holding answers without knoweldge, and to meet this technology with the cura  knowledge production demands, and not with avoidance: to meet technology with the thauma of inhabiting the tension, rather than striving to resolve it. This resistance operates at three scales. It is epistemic: declining to take the archive’s distribution of the plausible as if it were the shape of the world. It is political: insisting on what that archive makes unlikely, on the disparate, on what cannot be integrated. And it is juridical: refusing to grant the sovereign the standing, that is refusing to grant co-authorship, personhood, rights, as either would naturalize its position. The first two are practices. The third is a decision, and it returns us to the questions with which this essay began, which turn out to matter after all, though not quite for the reasons usually given.

This is a written version of a paper delivered at the 2026 Critical Legal Conference, University of Westminster, in the stream “Law and Process”. Many thanks for the questions from the audience, which have prompted revisions and improvements to the original paper.