The AI Covenant

The Typo

I saw an article in The Wall Street Journal about people who write for a living. Something strange was happening across all of them. A copywriter has started dropping in phrases like "hey yo, for real" and stacking up exclamation points, on purpose, because sounding polished now reads as sounding synthetic. They say it feels icky. They do it anyway. A blogger litters his posts with references no AI would think to reach for, obscure lines from The Office, because a machine will not naturally go there and he needs proof that he did. A financial coordinator leaves an accidental typo in when he edits. A tech founder has started replacing his em dashes with two small dashes instead, and calls the habit his artisanal craftsmanship. None of these people are bad writers. They are good writers, deliberately damaging their own work, because a clean sentence has started to look like evidence against them.

Rest with that image for a moment before you make your claim about artificial intelligence. People are roughing up their own prose to prove a human being made it. The flaw has become the credential. The very credential a person spent years and thousands of dollars earning. And the typo — the exact thing your teacher's red pen exists to destroy — has become evidence of humanity.

Everyone in that list thinks they are having an argument about a technology. They are not. Before we can say a technology is immoral, we have to say what moral claim it has violated. Most of us skip that step. We go straight to the verdict. Good or bad. Cheating or not cheating. Real or fake. We have quietly started confusing where a thing came from with whether it is good. We have begun confusing provenance with virtue.

A wrong question, asked forcefully and repeatedly enough, can occupy a whole society for decades. The question is not, is AI good or bad? Mine is narrower: what are we trying to preserve about the human being here?

If This Is a Moral Crisis, Let Us Actually Do Moral Reasoning

If artificial intelligence really is a moral crisis — and I believe some version of it is — then the honest move is not a slogan. It is not simply an emotional reaction. It is moral reasoning, the actual discipline of it.

While studying at Union Theological Seminary in New York, I learned about an old tradition in moral theology called casuistry, a method closely associated with the Jesuits. Casuistry refuses to solve a hard question with a slogan. It looks at the particular case in front of you, because morality lives in particulars, not categories.

The classical structure asks three things of any act. The object: what is actually being done? The intention: why is the person doing it? The circumstances: under what conditions is the act occurring? That framework gives us the method, but it was never designed with artificial intelligence in mind. For the case in front of us, I want to add three more questions. Covenant: what does this activity implicitly promise? Epistemic ownership: who actually possesses the understanding behind the act? Responsibility: who answers for the result?

If you cannot locate the exact rung where legitimate assistance turned into cheating, you do not actually have a clear moral objection to AI. You have a feeling about it, and feelings deserve honesty and interrogation, not the dignity of a verdict.

Whose Thinking Is This

We need to separate two things that AI has collapsed into one another in the public imagination: producing words and possessing knowledge.

There is a difference between epistemic authority — why anyone should trust that I know something — and epistemic ownership — whether I can actually claim the understanding as mine. Can you say: I understand this. I believe this. I can explain it. I can defend it. I can revise it when somebody pushes back. I accept responsibility if it turns out to be wrong. If you can say all of that, substantial epistemic ownership still belongs to you, even if AI was part of the room where the thinking happened.

The countercase: somebody asks AI to explain Kierkegaard, builds an argument on top of it, submits the argument, and cannot explain a word of it when a professor pushes back. The words bear their name. The knowledge does not. That is theft of a strange kind, a kind that robs the thief themselves.

The words can belong to you while the understanding does not. And the reverse is also true. Human beings have always thought in community. We think with teachers, editors, books, libraries, colleagues, parents, sermons we half remember, search engines, the whole slow accumulation we call culture. AI changes the power and proximity of that assistance. It does not change the fact that none of us has ever thought entirely alone.

So the question is not, did a machine touch this? The question is, did the machine extend the thinking I was already doing, or replace the thinking I was supposed to do myself?

"AI Is Just a Tool" Is Not Good Enough

Somebody, in nearly every conversation about this, will eventually say: AI is just a tool. And they are right. Yet it tells us almost nothing.

Steroids are a tool. Running shoes are a tool. A calculator is a tool. Photoshop is a tool. A motorcycle is a tool. Calling something a tool does not tell us whether using it is right or wrong. Every one of those tools is perfectly acceptable in one setting and cheating in another.

A calculator in an accounting department is simply how the work gets done. The same calculator, used during an exam built to measure whether the student can do the math, defeats the whole purpose of the exam. A motorcycle to get to work is just transportation. A motorcycle ridden during a marathon is fraud, and everybody already knows it without needing a lecture. The machine did not change between those two cases. The covenant did.

Every activity carries an implicit agreement about what a human being is supposed to contribute to it: the covenant of the activity. Usually it is unwritten, and we only notice it exists the moment somebody breaks it.

The Washing Machine and the Marathon

Nobody stands over a washing machine and says, your grandmother scrubbed those clothes by hand, and you pushed a button, and you cheated. The point of laundry was never the scrubbing. The point was clean clothes. The labor was only ever a means to that end, and when technology removes the labor and keeps the end, we call it progress.

But nobody accepts "I covered twenty-six point two miles" if the miles were covered on a motorcycle. Why? Because arriving was never the point of a marathon. The running is the point. The labor is not a means to the achievement. The labor is the achievement. Remove it, and there is nothing left to admire.

That difference gives us a taxonomy:

But that is only half the truth, and the other half complicates it immediately. AI can also challenge a student, interrogate their assumptions, demand they explain themselves, generate counterarguments they have to answer. Used that way, it increases formation instead of erasing it. Which is why the only honest governing question, in a classroom or a newsroom or a sanctuary, is this: what disappears from the human being when this labor disappears?

Labor as Identity

Notice our language. We do not usually say, he writes. We say, he is a writer. Not, she practices medicine, but, she is a physician. Our language gives us away. Somewhere along the way, we transformed labor into ontology. I do, therefore I am.

There is a cultural equation running underneath much of what we believe about work: labor produces competence, competence produces identity, identity produces status, and somewhere along the way we learned to confuse economic value with social worth. AI threatens several links in that chain at once, and that is why the fear it produces feels so much larger than "some tasks will be automated." It is not a fear about tasks. It is a fear about the self.

When a technology lets someone arrive quickly at a destination that took you twenty years to reach, it does not merely threaten your future income. It can feel like it devalues your past suffering. The anger deserves to be taken seriously.

But casuistry demands we separate two things that principle has fused together. There is a difference between saying, "The struggle produces something intrinsically necessary," and saying, "I struggled, therefore you must struggle too." The first is a claim about formation. The second is not a moral claim at all. It is grief dressed up as moral conviction, and we owe each other the honesty to tell the difference.

Technological disruption is a moment of moral disclosure, not moral creation. AI did not invent the lie that your worth equals your output. It only turned the lights on in the room where that lie has been living the whole time.

Sovereign Delegation

The smartphone arrived as convenience. It did not create distraction. It industrialized the competition for our attention. Can I govern my attention when the systems around me are designed to be extraordinarily good at capturing it?

AI takes the problem one level deeper. Can I govern my thinking when the systems around me are extraordinarily good at assisting it, and increasingly performing parts of it for me?

And beneath both sits a third question, the one we are heading toward as these systems mature: can I govern my identity when machines become extraordinarily good at reflecting me, predicting me, persuading me, even performing versions of me?

Call it sovereign delegation: the capacity to consciously determine what should remain mine, what may be delegated, what should be augmented, and what must never be surrendered.

Sovereignty does not mean doing everything yourself. That is not sovereignty. That is exhaustion dressed up as integrity. Sovereignty means retaining meaningful authority over the act of delegation itself, so that when you hand something to a machine, you handed it, rather than simply drifted into losing it.

Some things can be delegated without much moral weight attached. Spellcheck. Transcription. Formatting. Information retrieval. Administrative work. Other things require far greater caution: judgment, belief, moral responsibility, meaning, identity, relationship, formation, and any high-consequence decision that will still belong to you long after the machine has moved on to someone else's request.

It is not merely that AI will think for us. It is that we may slowly lose the ability to recognize which thinking was supposed to remain ours. That loss will not announce itself. It will feel, at every step, like convenience.

We Need New Covenants

Every covenant we currently operate under — in the university, at work, in the studio, in the pulpit — assumed overwhelmingly human labor and human cognition on the other side of every exchange. That assumption is breaking, quietly, in institution after institution, and most have not yet noticed the ground has moved.

Not one AI rule for everybody. Multiple covenants, because different human activities exist for different purposes, and a single blanket policy will always fail somebody.

The Four Foundations of an AI Covenant

Not twenty commandments. Nobody remembers twenty commandments, and moral architecture only survives if it is memorable enough to carry into the next hard decision.

What Kind of Human Must Emerge

The emerging human will live alongside nonhuman intelligence. Not beneath it. Not pretending it does not exist. Not by refusing to touch it. Not necessarily competing against it in every domain where it now outperforms us, because tying human worth to staying ahead of the machine leaves human dignity hostage to every new advance.

What kind of human being must emerge in order to remain human beside machines capable of performing an increasing share of what we once used to define ourselves by?

I do not have a complete answer, and I do not trust anyone who claims to. But I know some of what belongs in it. A human being capable of discernment and of sovereign delegation. A human being who understands the difference between information and knowledge, and does not confuse having access to the first with possessing the second. A human being capable of epistemic ownership, who can stand behind what bears their name. A human being whose identity is not reducible to their occupation, so that when the occupation changes, or disappears, they do not disappear with it. A human being capable of relationship, who understands that relationship asks something more of us than simulation because it requires vulnerability, obligation, and risk. A human being whose dignity does not depend on outperforming machines.

That list stays open, not closed. Finish it, and you start worshiping the list instead of living the questions inside it.

Return to the Typo

A typo does not prove judgment. It does not prove understanding, formation, responsibility, conscience, experience, meaning, love, suffering, or wisdom. It proves only that there is a typo. We built an entire anxious ritual around the wrong evidence.

We began by asking, how can I prove a human made this? That was never the important question. The important question, the one this essay has been circling from its first page, is this: what are we trying to preserve about the human being here?

Casuistry can sort the cases, but theology has to ground the worth. Human worth must precede human productivity. Not follow it. Not depend on it. Not wait to be earned by it.

We will need new covenants. Between student and university. Between worker and employer. Between artist and audience. Between pastor and congregation. Between citizen and institution. And before any of them, a covenant with ourselves.

What will we delegate? What will we preserve? What will we keep practicing even when a machine can do it better? What will we remain responsible for? What forms of difficulty will we endure because the difficulty is still forming us? What will we refuse to surrender, not because a machine cannot do it, but because doing it is doing something to us?

Those answers will change. The obligation to ask will not.

And beneath every one of those covenants sits an older question, one that was never actually about artificial intelligence at all.

If what I do is no longer sufficient to tell me who I am, who must I become?