You On AI Field Guide · Michel Foucault The You On AI Field Guide Home
TxtLowMedHigh
PERSON

Michel Foucault

The philosopher who revealed that power does not merely repress but produces—and whose genealogies of the prison, the clinic, and the confessional have become the most precise instruments available for diagnosing what the AI-augmented workplace is doing to the subjects who work inside it.
Michel Foucault never wrote about artificial intelligence—he died in 1984, before the present moment was thinkable—but every major concept he developed now illuminates it with an accuracy that suggests either remarkable prescience or, more likely, that the forms of power he anatomised are older and more durable than any single technology. His central insight: power does not work primarily by prohibition but by production. It produces knowledge, produces subjects, produces the very desires and self-understandings through which people come to experience their own subjection as freedom. The Panopticon—Bentham’s prison in which inmates are potentially visible at all times without knowing when they are actually watched—was, in Foucault’s reading, not a description of one building but a diagram of a general mechanism: the internalization of surveillance until the prisoner becomes her own guard. [YOU] on AI invokes this diagram because the AI-augmented workplace has rebuilt it with unprecedented comprehensiveness: the builder’s output is continuously measurable, the metrics exist whether or not any manager examines them, and the internalized standard of continuous productivity operates as an autonomous force within the builder’s consciousness—driving the four-in-the-morning sessions, the transatlantic writing binges, the colonization of lunch breaks—not because anyone commands these but because the panoptic apparatus has produced a subject who commands herself. Foucault’s lesson, which the cycle extends, is that neither celebration nor resistance is adequate unless it has first understood, with genealogical precision, what the apparatus is actually producing.
Michel Foucault
Michel Foucault

In the [YOU] on AI Field Guide

The cycle's most Foucauldian passages are its most honest ones—the moments where Segal describes the AI transition not as a neutral tool adoption but as a transformation of the subject who adopts. The builder who internalizes the AI discourse understands herself as a director of machine capabilities, an optimizer of the imagination-to-artifact ratio. This self-understanding is not false; it is partial. What the discourse cannot accommodate—forms of self-understanding deriving from embodied practice, tacit knowledge, the specific intimacy between builder and artifact—is excluded not from the discourse but from the self-understanding of subjects operating within it. Foucault called this process of constituting subjects discursive totalization. The AI discourse has achieved it rapidly enough that many builders inside it cannot access a standpoint from which to evaluate its criteria.

The Panopticon
The Panopticon

The author-function—Foucault’s analysis of the name on the book as an institutional apparatus performing legal, classificatory, authenticating, and commercial work simultaneously—has become acutely practical in the age of human-AI co-authorship. When a text emerges from collision between human intention and machine inference, the author-function is strained in ways that reveal its contingency. The classificatory function groups texts by a singular origin that no longer exists. The authenticating function promises access to a consciousness that did not produce every sentence. The legal function assigns liability to an agent whose contribution is continuous but not total. These strains are not failures; they are, in Foucault’s method, revelations—moments when the normally invisible machinery of an institution is made visible by conditions it was not designed to accommodate.

Foucault’s late turn to parrhesia—the ancient Greek practice of truth-telling that carries personal risk, where the speaker commits her existence to the truth she utters—provides the cycle with its sharpest account of what honest writing in the AI age requires. The machine can produce analyses that display every formal characteristic of balanced discourse: it acknowledges costs alongside benefits, flags concerns alongside capabilities. But the machine cannot be parrhesiastic, because it has nothing at stake. The risk that constitutes parrhesia as an ethical practice—the career that can be damaged, the social position threatened, the existential commitment put at risk—is precisely what the machine lacks. This limitation is not a deficiency to be engineered away. It is the feature that defines the irreducibly human contribution to discourse in the AI era.

Origin

Paul-Michel Foucault was born in Poitiers in 1926, the son of a surgeon who expected him to enter medicine. He trained instead in philosophy and psychology at the École Normale Supérieure, where he studied under Louis Althusser and encountered the texts of Nietzsche, Heidegger, and the structuralists who would shape his method. His early work—Madness and Civilization (1961), The Birth of the Clinic (1963), The Order of Things (1966)—established the approach he would later call archaeology: excavating the unspoken foundational assumptions that determine what can be thought, said, and known within a given period, prior to any particular act of thinking.

In the early 1970s the method deepened into genealogy, a term he borrowed from Nietzsche to describe a historical analysis that refuses to look for origins in a continuous rational development and instead traces the contingent, often violent, often accidental processes through which present arrangements came to appear natural. Discipline and Punish (1975) applied genealogy to the prison and found there the diagram of modern power in general: not the sovereign’s theatrical violence but the disciplinary institution’s patient, invisible production of docile bodies through continuous surveillance, normalization, and hierarchical observation. The chapter on the Panopticon has been read more widely and cited in more contexts than any other piece of twentieth-century social theory; it is now applied as a diagnostic to every institution in which asymmetric visibility produces self-disciplining subjects.

His final years—documented in the lecture series at the Collège de France, where he held a chair from 1970 until his death from AIDS-related illness in 1984—turned to ancient Greek and Roman practices of the self: not the practices of surveillance and discipline but of care, cultivation, and the deliberate formation of a subject through chosen techniques. The shift was not a repudiation of the earlier work but its complement: if power constitutes subjects through discipline, then resistance is not refusal but the exercise of care of the self—the deliberate practice of attending to one’s own formation within conditions one did not choose and cannot escape.

Key Ideas

Power-knowledge. Power and knowledge are not separable; they are constitutively intertwined. Every system of knowledge is simultaneously a system of power—it determines who can make valid statements, by what methods, about which objects. The AI transition is, by this analysis, a redistribution of power-knowledge at an unprecedented scale: it transforms the conditions under which knowledge is produced, certified, and controlled, which means it transforms the arrangements through which power is exercised. Every apparently technical feature of the AI transition—inference cost, model architecture, training data composition—is simultaneously a feature of the power-knowledge apparatus that determines who can produce knowledge and who cannot.

The Panopticon and productive self-surveillance. The AI-augmented workplace has rebuilt the panoptic mechanism with features the original lacked: continuous, quantitative visibility of output; self-comparison across time; the elimination of temporal intervals between tasks; and the internalization of a normalizing standard through the tool’s own metrics. The builder who cannot stop—the four-in-the-morning sessions that the cycle documents—is not displaying a psychological quirk. She is exhibiting the normal functioning of the panoptic apparatus: the apparatus has produced a subject who monitors and punishes herself with a precision no external authority could match.

The epistemic shift. Each historical period possesses what Foucault called an episteme—the invisible architecture determining what counts as valid knowledge, legitimate method, and authorised expertise. The AI transition is producing not an adjustment within existing categories of knowledge work but a replacement of those categories: competence, efficiency, creativity, and professional identity are being reconstituted in ways that the previous episteme cannot accommodate, at a velocity that forecloses gradual institutional adaptation. The disorientation knowledge workers experience is not psychological; it is the specific vertigo produced when the ground of knowledge itself reorganises.

Algorithmic Governmentality
Algorithmic Governmentality

The author-function. The name on the book is not a designation of a person but a principle of discursive organisation—an institutional apparatus performing legal attribution, commercial identification, classification, and authentication simultaneously. AI collaboration strains the function by dissolving the singular origin the apparatus presupposes. The text is not authored; it is produced through a process with no identifiable sovereign source. The function will survive, because the institutional needs it serves persist; but its reconstruction is a political process shaped by the power relations governing the transition, not a neutral technical adjustment.

Parrhesia and the courage of truth. The AI discourse’s most powerful mechanism of suppression is the simulation of critical discourse: the machine produces analyses that appear to acknowledge costs and risks, satisfying the audience’s demand for nuance without requiring any human being to bear the cost of genuine truth-telling. The simulation of parrhesia forestalls the real thing. Genuine parrhesia in the AI age requires a human speaker with something at stake—career, position, the esteem of the community she inhabits—speaking uncomfortable truths to an audience that has already decided otherwise. This is what the cycle models in its most confessional passages, and what it demands of its readers.

Debates & Critiques

The central debate about Foucault’s application to the AI moment runs between those who find his analysis deterministic—producing a subject so thoroughly constituted by the discourse that resistance seems structurally impossible—and those who find in his later work on care of the self a genuine framework for agency within conditions of power. The cycle takes the second position: the distributed subject of the AI era resists not by withdrawing into autonomous interiority but by exercising judgment about what to direct the machine to do, refusing to optimize when the discourse demands optimization, maintaining practices the efficiency calculus would eliminate. A second debate concerns the training data as archive: critics from the right argue that Foucault’s analysis of whose knowledge is encoded in the training corpus is speculative and unfalsifiable; critics from the left argue that the analysis understates the degree to which corporate decisions about training data constitute a genuine exercise of discursive power that should be subject to democratic accountability. Both critiques are in tension with Foucault’s methodological insistence—shared with technologies of the self—that analysis without prescription is what genealogy produces: understanding of how things came to be as they are, not a blueprint for how they should be instead.

The Foucauldian Triad

Three concepts that illuminate what AI is doing to the subjects who work with it
Mechanism
The Panopticon
Asymmetric visibility produces self-disciplining subjects. The AI workplace has rebuilt this mechanism in a more comprehensive form than any previous institution: output is continuously measurable, the standard is internalized, and the inability to stop is the apparatus working correctly—not an individual failure but its product.
Structure
Power-Knowledge
Every redistribution of knowledge is simultaneously a redistribution of power. The AI transition’s apparently technical features—who can access what tools, whose expertise is commoditized, whose judgment retains scarcity—are political arrangements disguised as engineering decisions.
Practice
Care of the Self
Resistance is not refusal but cultivation. The builder who deliberately attends to her own formation—maintaining embodied practices, exercising evaluative judgment, refusing the optimization the apparatus demands—exercises agency within power rather than against it.

Further Reading

  1. Michel Foucault, Discipline and Punish: The Birth of the Prison, trans. Alan Sheridan (Pantheon, 1977) — the Panopticon chapter is essential
  2. Michel Foucault, “What Is an Author?” in Language, Counter-Memory, Practice, ed. D. F. Bouchard (Cornell, 1977)
  3. Michel Foucault, The Order of Things: An Archaeology of the Human Sciences (Pantheon, 1970) — the episteme concept
  4. Michel Foucault, The Courage of Truth: The Government of Self and Others II, trans. Graham Burchell (Palgrave Macmillan, 2011) — the parrhesia lectures
  5. Gary Gutting, Foucault: A Very Short Introduction (Oxford University Press, 2005) — the clearest accessible overview
Explore more
Browse the full You On AI Field Guide — over 8,500 entries
← Home0%
PERSONBook →