Servitude as a Service: Reading Kant in the Age of Artificial Intelligence

Each semester, I ask my students to read Immanuel Kant’s 1784 essay, “What is Enlightenment?” We begin with the essay’s historical context: the Age of Enlightenment, the republic of letters, despotism. While that is important, I must admit that I — and most certainly the students — enjoy the class more when our discussion turns away from the past and towards our own moment.¹ One sentence in the essay’s third paragraph consistently catches the attention of my most discerning students: “Rules and formulas, those mechanical aids to the rational use, or rather misuse, of his natural gifts, are the shackles of a permanent immaturity.” We talk about ideology, dogma, and prescribed beliefs like the scourge that is partisan politics. When we talk about the tools that discourage independent thought, someone inevitably says: “Like ChatGPT?”

Yes. Exactly like ChatGPT.

Kant warned about “mechanical aids” that trap us in “permanent immaturity” — tools that seemingly help us think but actually prevent us from thinking. Two and a half centuries later, we subscribe to them monthly. Servitude as a service: that’s what AI platforms are selling us.

The Child and the Beast

Kant defines “enlightenment” as intellectual maturity. It is, in his words, “the ability to use one’s own understanding without guidance from another.” We might call this today “independent thinking.” It is the application of our natural intellectual faculties. It is thinking and reasoning for oneself, freely and critically.

His metaphors for those who cannot think independently are pointed. The unenlightened are childlike, “immature,” and in need of “supervision.” But they are also beastlike: “domestic livestock,” “docile creatures,” yoked and harnessed. At best, they are toddlers. At worst, they are sheep. Learning to think for oneself, Kant suggests, is like a child or a foal learning to walk. They figure it out eventually, but not before falling a few times and eventually developing the courage to continue alone.

Despite the deameaning tone, his metaphor is clear: to be immature and unenlightened is an intellectual dependence and servitude.

The Offloading of Thought

To write is to think. And when we offload our writing to a tool, we are doing less thinking. But reading, too, is thinking. As I argued elsewhere, those who even casually use artificial intelligence have no clue they are reading less. Those who deliberately abuse it, reading less is the whole point. But because the link between intelligence and reading is well-established, you can see how the collapse of reading might be a problem. A dependence on artificial intelligence is a crime against rationality.

Heavy AI users offload the heavy cognitive lifting to a “mechanical aid,” an act that will have a detrimental impact on measurable intelligence — that is unless we simply lower our standards of evaluation, which is already happening. Their brains aren’t getting the reps they need to stay fit, so to speak. They aren’t exercising those muscles.

Some applications of artificial intelligence will prove valuable, like those in health sciences, research, and specialized technical fields. Would I like AI to copyedit this essay and compile a list of tags? Yes, please! But the most pernicious uses of AI, in Kant’s framework, are the ones marketed as the most helpful: tools that let us offload cognitively demanding tasks like reading, writing, and thinking.

The Guardians and Their Livestock

If the unenlightened are children and animals, then naturally there are adults and masters in the room. Kant calls these supervisor-like figures “guardians.” He conjures up an image that students recognize as caricatures of the “deep state,” the man in the smoky room, the puppet-master controlling narratives for their own interests, the partisan media.

But in the era of artificial intelligence, the guardians who benefit from our ignorance are the Big Tech oligarchs.

There’s a direct financial relationship between users and AI services. Many AI services charge subscription fees. Soon, I suspect, they’ll be “free,” generating income instead through data collection and ads. Like social media platforms, AI services will sell your queries and data to third parties, marketers, and advertisers. Even AI platforms will become enshittified, if they haven’t already.

But the situation becomes increasingly dire the more we consider Big Tech’s cozy relationship with the political class.

The DeepSeek Example

To highlight the problem, consider DeepSeek, the Chinese AI platform that briefly made waves as a free ChatGPT alternative. Ask DeepSeek about Tiananmen Square, and it refuses to discuss the topic: “I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses.”

This is everything you need to know about how ruling-party ideology gets embedded in these supposedly “intelligent” systems where information that might undermine authority is deemed harmful.

AI tools are not apolitical. They’re not objective fountains of information. Their outputs are guided and manipulated, if not directly then by its training data. They have biases much like the Encyclopédie during the Enlightenment did, whose articles condemning slavery, attacking religion as superstition, and questioning the divine right of kings got its publication and authors in trouble with the ruling class. The King’s censors restricted the book for “destroying royal authority, fomenting a spirit of independence and revolt, and… laying the groundwork for error, for the corruption of morals, and for irreligion and atheism.”²

If AI outputs can be manipulated by political pressure in China, they can be manipulated in the United States. Though nominally more open, many see the U.S. sliding toward authoritarianism. And if that’s the case, the Trump administration’s cozy relationship with Big Tech should concern us.

AI outputs can be and are fine-tuned by their developers to align with their political leanings. Take Grok, Elon Musk’s AI service integrated into X. Musk, designed Grok to be an ideological counterweight to what he believed was a left-leaning and politically correct ChatGPT. Grok now consistently provides outputs that align with Musk’s own socio-political ideas — whatever they may be at the moment.

A Descent into Ignorance?

The general misunderstanding that artificial intelligence is an objective system, god-like in its omniscience, makes the problem worse. AI outputs cannot be trusted as objective truth.

AI tools are not our “guardians.” Overusing them, especially for higher-order thinking skills like reasoning, analysis, reading, and writing (to say nothing about creativity!), prevents users from exercising and cultivating their natural intellectual abilities. It leads to servitude and dependence. This is what Kant warned us about nearly 250 years ago.

If Kant remarked that he lived in an “age of enlightenment,” what would he say about our age? An age in which roughly half of US adults use AI at least once a day… and rising?³ An age in which AI is being embedded everywhere — email replies, search results, text message summaries, classrooms? Our ignorance is becoming inescapable.

This isn’t just about individual enlightenment. Kant was concerned with the larger public’s intellectual health — our collective ability to think and reason independently was necessary for a functioning society. He believed in intellectual progress, that humanity could emerge from its self-imposed immaturity. This required courage and effort.

But what happens when we’ve built systems that make immaturity easier than enlightenment? When thinking becomes optional, and the path of least resistance is one that leads us away from our natural ability to think and reason? When our servitude is offered to us as a service (SaaS)?

The Age of Information may be giving way to the Age of Ignorance. A dependency on artificial intelligence is a wonderful way to create natural stupidity. The irony would be almost funny if the stakes weren’t so damn high.