Foucault at 100: The Circle of Truth and Power in the Age of AI

A basic, necessarily simplified introduction to one aspect of Michel Foucault

Executive summary: the questions we need to ask now

How does AI shape the way we produce truth?

How do we collectively choose what we accept as true, and how do those choices reshape power relations?

How much of that choice is technology already making for us?

These questions become difficult to ask when our mental model treats truth as universal, simply waiting to be discovered. We look for the correct answer and overlook how an answer becomes intelligible, credible and authoritative in the first place.

In “Truth and Power”, Foucault gives us a way to examine that process. Power shapes the production of truth, and truth, in turn, shapes power. As AI changes how we form our understanding of the world, this circle becomes essential to understanding what is happening to us.

1. Why we need Foucault now

Michel Foucault would have turned one hundred today, 15 October 2026.

For his centenary, I want to share one idea that changed how I understand the world: truth and power form a circle. Understanding that circle matters when we discuss science, when we decide whom to believe and what truth means, and when we build technologies that increasingly organise our social lives.

This is a modest tribute and a deliberately simplified introduction. I will take shortcuts.

My goal is to remove some of the obstacles that usually make Foucault's insight seem shocking, absurd or impossible to accept.

Ideas can take generations to become common sense. But we cannot afford to wait another two hundred years for Foucault's ideas to become widely understood. We need to update our mental model of truth, science, technology and power today. We can start right here, as you read this article.

2. Doesn't science give truth a universal foundation?

Here is the most common objection: But science is universal. Doesn't that give truth a universal foundation?

It deserves an answer.

We often imagine scientific knowledge as a tower: axioms at the bottom, logical deductions above them, reliable conclusions at the top. Empirical science also depends on observation and experiment, of course. But let us start with mathematics, where the promise of certainty seems strongest.

What does an axiom mean?

Consider a familiar principle of Euclidean geometry: through any two distinct points, exactly one straight line can be drawn.

But what is a point? What is a line? What do through, two and exactly one mean?

We can offer definitions. But each definition uses other terms. Ask what those terms mean and the problem returns.

Primitive terms still require understanding

Calling a term primitive does not remove the need for a starting point. A formal system can leave some terms undefined and specify their relationships through axioms, as Hilbert does with points, lines and planes in The Foundations of Geometry. This establishes where the formal development begins. It does not make that beginning intelligible by itself.

Take the notation {} for the empty set. Someone still has to understand what this notation means and how it is used. Explaining it through further symbols or rules only shifts the question: how do we understand those starting points?

How do we know they are a solid foundation for what we will call universal truth?

Even if we treat the symbols purely as signs to manipulate, we must still recognise them and understand what it means to follow the specified rules. We have chosen where to begin; we have not produced a beginning that requires no prior understanding.

The human starting point

For scientific knowledge to develop, human understanding has to begin somewhere.

One person shows another a diagram, gives an example, corrects a mistake. Eventually, they can reason together. This is where shared understanding begins.

The shared understanding might begin with an intuition; it can become much more precise through training and formalisation. But another definition, on its own, cannot guarantee that they will understand one another.

Even at this apparently pure foundation, there are people.

Truth depends on human beings who have learned to recognise something together.

The familiar argument for universality has now fallen apart. We were promised an impersonal foundation: axioms, then deductions, then truth. But the axioms do not explain themselves. Their intelligibility depends on practices shared among people.

3. Why shared understanding is political

But a shared understanding is social. Why call it political?

Shaping how people come to understand something together is a political act because, by definition, politics concerns how we organise our lives together.

A professor stands at the front of a room. Students sit facing them. One person writes the axiom; the others are expected to be quiet. There is a timetable, an examination, an institution that awards qualifications. Someone decides who may enter that room.

The participants have bodies. They occupy different positions, have different opportunities and depend on others in different ways. The conditions under which they acquire knowledge are material.

The professor's position does not make an invalid proof valid. But it matters enormously to whose reasoning gets heard, which concepts become familiar, and who acquires the authority to produce the next explanation.

The relationship is very different when you learn with an AI assistant.

Foucault's technologies of power

This is my way into Foucault's question about power. His historical investigations draw our attention to how bodies, observation, classification and authority are organised in the production of knowledge.

In “Technologies of the Self”, Foucault uses technologies to describe organised practices through which people act on things, signs, others and themselves. His technologies of power concern the ways people's conduct is directed and organised. An examination, a hierarchy or an arrangement of bodies in a classroom can be analysed in these terms.

Organising how knowledge is shared is an exercise of power: it is how truth is created and reality is constructed.

4. Doesn't peer review establish universal truth?

But scientists check one another's work. Doesn't that solve the problem?

Peer review helps identify errors, but it does not guarantee unanimity. Researchers may disagree about a new concept, a proof or what the evidence establishes. Some disputes last for decades. The understanding through which a result becomes accepted as true is worked out collectively. Checking one another's work makes this social foundation visible.

5. The rocket: when knowledge becomes power

But the rocket flies. Doesn't that show that science isn't arbitrary?

Start a hundred years earlier.

Imagine two groups of mathematicians disagreeing over a proposed theorem and the concepts on which it rests. One regards the framework as sound and promising. The other challenges its formulation, its proof or its significance. Researchers review one another's work, but agreement is incomplete. Both schools teach, publish and train students.

Now move forward a century. Engineers use the mathematical framework developed by one of those schools to help design a rocket.

The rocket flies.

Its launch demonstrates a real capacity to predict and act on the world. It also gives a company or a state new capabilities: communication, observation, access to space and military reach.

Those capabilities create power. Funding flows towards the institutions associated with the achievement. Their researchers gain prestige. Their methods spread through laboratories and teaching. Their account of the field becomes the standard against which new work is assessed.

The once-disputed ideas of one school are now taught and accepted as true.

Truth is created only when it increases the power of those who claim it. Truth is nothing but power.

The rocket has helped transform a disputed body of knowledge into an authoritative one. The power produced by its success supports the institutions that establish and transmit that knowledge.

Knowledge makes the rocket possible. The rocket creates power. That power helps establish the knowledge as truth.

6. How AI changes the production of truth

We can now return to the questions we began with.

• How does AI shape the way we produce truth when it can present each of us with a different account of the same world, each presented as knowledge?
• What happens when AI can tailor persuasive messages to individuals?
• How do we collectively choose what we accept as true, and how do those choices change who can speak, decide and act?
• How much of that choice is technology already making for us?
• Who decides which sources an AI treats as credible, which disagreements it makes visible and which assumptions it presents as settled?
• What happens when its answers become the starting point for education, journalism or decisions that affect other people?
• Who gains power through these choices, and how can that power be challenged?

7. Why our old mental model gets in the way

If we imagine truth as something universal that technology should retrieve for us, these choices become hard to see. We judge the answer while leaving the process that gives it authority largely unexamined.

We expect disagreement to disappear once the correct facts have been supplied. We struggle to understand why people inhabit different accounts of reality, or why a persuasive account can reorganise power before its claims have been seriously tested.

We cannot understand this transformation while treating the production of truth as though it had no politics.

Foucault gives us the questions we need to ask about how that production is organised, who participates in it and what it makes possible.

8. We cannot wait another two hundred years

The danger concerns us all. Technologies can reshape what people accept as true while our inherited mental model leaves us poorly equipped to recognise the power being exercised.

We risk handing over important decisions without even recognising that choices are being made.

Understanding the circle of truth and power helps us see the limits of that model and begin to imagine different ways of organising our lives together. We need to discuss how shared truths emerge, how authority is established and how those arrangements can be challenged.

Foucault would be one hundred today. A useful tribute would be to make these questions part of how we educate, build institutions and design the technologies now entering our lives.

Truth and power form a circle. We cannot afford another two hundred years to learn how to recognise it.

Sources and scope

Michel Foucault, “Truth and Power”, in Power/Knowledge: Selected Interviews and Other Writings, 1972–1977, edited by Colin Gordon, 1980, especially pp. 131–133. The link leads to the interview's English journal publication.

David Hilbert, The Foundations of Geometry, translated by E. J. Townsend, 1902, Chapter I, sections 1–2.

Gary Gutting and Johanna Oksala, “Michel Foucault”, Stanford Encyclopedia of Philosophy, especially the discussion of Discipline and Punish.

Michel Foucault, “Technologies of the Self”, seminar at the University of Vermont, 1982; published in Technologies of the Self: A Seminar with Michel Foucault, 1988, pp. 16–49.

Sandra C. Matz and colleagues, “The potential of generative AI for personalized persuasion at scale”, Scientific Reports 14, 4692 (2024).

The centenary date is confirmed by the Collège de France.

The two schools and the rocket form a thought experiment, not a documented historical episode. The political implications developed here extend beyond the cited persuasion research. “Two hundred years” expresses urgency, not a measured law about the diffusion of ideas.