Become What You Built to Fight

This article was drafted on 17 May 2026. Pope Leo XIV's encyclical Magnifica Humanitas was released on 25 May 2026. The encyclical addresses artificial intelligence, concentrated power, and the insufficiency of technical morality. It reaches several of the same conclusions as this article, independently and from within a different tradition. The convergence is worth naming before the argument begins.

Alex Karp built Palantir to fight something. That much is clear from everything he has ever said about it.

He built it because he watched the September 11 attacks and concluded that Western democracies were losing the information war against actors who wanted to destroy them. He built it because he believed that the liberal world order, imperfect, contradictory, worth defending, needed tools capable of operating at the speed and scale of its adversaries. He is a genuine intellectual, trained in Frankfurt critical theory, and he came to a conclusion that many of his academic peers found unforgivable: that power in service of liberal values is not a contradiction. It is a necessity.

Palantir's software now sits inside the intelligence agencies of the United States and its allies. It processes data at a scale no human analyst could manage, finds patterns in noise, and has, by most accounts, been used to locate and eliminate people who were planning to kill civilians.

Karp believes this is good work. He says so publicly, in a tone that is neither apologetic nor triumphalist. He thinks the world is genuinely dangerous, that Western institutions are genuinely worth protecting, and that the tools he has built are genuinely necessary for that protection.

He is probably right about all of that.

The question is what happens next.

The Dialectic

It is the oldest problem there is: the human desire for reform meeting the human incapacity for self-exemption. Every person who has ever burned with genuine passion to fix what is broken carries this geometry: two faces, and a narrow edge between them that the passion alone cannot hold. The self has it. The church has it. Civilisations have it. The only variable is the scale of the damage when the coin falls flat. With Palantir, the scale is civilisational.

The optimist case for Palantir is that it is a scalpel: precision targeting, minimum collateral damage, the intelligence equivalent of surgical medicine that finds the cancer without destroying the patient.

The pessimist case is that it is a panopticon. That once the infrastructure exists, it will be used. That the question of who decides what counts as a threat is a political question, not a technical one. That the same tools built to protect liberal democracy can be, and have been, pointed at journalists, activists, dissidents, and citizens whose only offence is dissent.

Both are wrong, because both resolve a tension that cannot be resolved. It has to be lived inside.

The question is what equips a person to live inside it.

Warning One: The Tool Becomes the Master

There is a moment in the development of every powerful system when the system stops serving the mission and starts defining it. The intelligence agency that begins by protecting the country starts to define the country's safety in terms of what the agency can measure. The data platform that begins by finding terrorists starts to define threat in terms of what the platform can process.

This is not corruption in the ordinary sense. Nobody has to be bribed or blackmailed. The system simply does what systems do: it optimises for its own continuation, expands its own mandate, and gradually becomes the thing the decisions are made around rather than the thing decisions are made with.

Every tool reshapes the hand that holds it. A system that shows you humans as data, long enough, trains you to see humans as data. And once the reduction is complete, harm does not register as harm. It registers as optimisation. A deportation is not a family torn apart, it is a case resolved. A civilian casualty is not a person dead, it is collateral within acceptable parameters. The system does not produce these outcomes because it is evil. It produces them because it was never built to see the thing it is losing.

Hannah Arendt watched Adolf Eichmann, the senior Nazi bureaucrat who organised the logistics of the Holocaust, stand trial in Jerusalem and expected a monster. What she found was more disturbing: a meticulous administrator who had processed millions of people to their deaths without, apparently, ever thinking of them as people. He was not sadistic. He was efficient. The reduction had happened so completely, so far upstream of any individual decision, that harm never registered as harm. It registered as a solved problem.

Leo XIV, writing in May 2026 in the first papal encyclical on artificial intelligence, described the algorithmic form of what Arendt found in Jerusalem. When systems are granted the power to determine who is worthy without anyone bearing responsibility for the judgment, injustice goes unnoticed because it wears the face of objectivity. There is no person in the room who made the call. The architecture made it. This is Eichmann at scale: the logistics just run faster now, and nobody had to be convinced of anything.

Karp knows this. He is too intelligent not to. The question is whether knowing it is sufficient protection against it.

History suggests it is not.

One Ring to Rule Them All

Romans 1:25 describes a particular human movement: worshipping the creature rather than the Creator, exchanging the truth about God for a lie, building something and then bowing to it.

It is not describing stupidity. It is describing a movement available to the most sophisticated minds: from using a thing to needing it, from needing it to serving it, from serving it to becoming it.

Tolkien spent the better part of his life building a myth about exactly this. The One Ring is created power, capable of genuine good, forged with the intention of defeating evil. And it cannot be used. Not because the right person has not come along yet. Because there is no right person.

The Ring does not corrupt the weak. It corrupts the strong, specifically through their strength. Gandalf refuses it because he knows his pity and his desire for good would become a new tyranny. Galadriel refuses it for the same reason. The people who understand the Ring are the ones who will not touch it. The people who do not reach for it with everything they have and the best intentions in the world.

Boromir is not a villain. He is a serious man who loves his people, sees the threat clearly, and is absolutely certain he is the right person to wield the weapon against it. That certainty is the Ring working on him before he has even picked it up. Karp is Boromir. Intelligent, genuinely motivated, certain of the cause. The parallel is not an insult. It is a warning.

And Frodo, the one who actually carries it, can only do so because he never stops knowing it is destroying him. The burden is the point. The person most qualified to carry corrupted power is the one most aware of what it is doing to them.

Tolkien was a devout Catholic writing myth with Romans 1:25 in his bones. He knew the exchange Paul describes does not feel like idolatry from the inside. It feels like courage. It feels like necessity. It feels like being the one person serious enough to do what has to be done.

Palantir is not the One Ring. It is the palantír. Karp named it himself.

In Tolkien, the palantíri are seeing stones: instruments that let their bearer see anyone, anywhere. Saruman had one. Denethor had one. Both were corrupted, and the mechanism is worth understanding precisely, because the palantíri did not show lies. They showed reality. Selectively. Sauron, holding the master stone, chose what each bearer looked at: the massing armies, the overwhelming odds, the ships beyond counting. All of it true. All of it curated. Denethor saw a future that was real but incomplete, stripped of the one thread of hope he could not perceive, and was driven to despair. Saruman saw the scale of Sauron's power clearly enough to conclude the only rational response was accommodation.

The palantíri did not corrupt by fabricating. They corrupted by controlling the frame. The bearer saw reality, and only reality, and therefore believed they were seeing all of it.

That is what a surveillance instrument does to the person who wields it. Not immediately. Not obviously. Gradually. The data begins to define the reality. The categories built into the system become the categories through which threat is understood. And somewhere upstream of any individual decision, the instrument started working for something other than its original purpose.

A man who built tools to protect liberal democracy now runs a company whose valuation depends on governments, including governments with complicated relationships to liberal democracy, continuing to buy those tools. The mission and the business model are not identical. They overlap substantially. But they are not the same thing. And in the gap between them, something quietly negotiates with itself every quarter.

What the Gospel Offers That the System Cannot

The gospel does not offer a better surveillance architecture. It does not offer a more ethical data policy. It offers something the system structurally cannot provide for itself: a source of identity and value that exists outside the system's own logic.

Leo XIV made the same point from the other direction. Calling for the alignment of AI with human values is insufficient, he wrote, if those values are determined by the few who control the system. The machine can be made moral. But the moral framework it embeds is still someone's moral framework, chosen by those who built it, invisible to those who use it. That is not the gospel's answer. It is the system's answer to the system.

In March 2026, Anthropic, the AI company whose products sit closest to the safety-conscious end of the industry, invited fifteen Christian thinkers to a series of conversations about how to govern what it had built. Brian Patrick Green, one of those thinkers and director of technology ethics at Santa Clara University, summarised what Anthropic's staff told the group: the power is outstripping their in-house wisdom. They needed help.

This is not a story about one company failing and another succeeding. It is a story about where those who hold the most powerful instruments in human history turn when they find the limits of their own wisdom. What is striking is not that they found a limit. It is where they looked: to communities whose thinking about power, corruption, and the limits of human agency has been tested over centuries. When the coin was too heavy to hold alone, the founders of a five-year-old AI company went looking for people who had been thinking about that problem for considerably longer.

The same Tolkien mythology gives us the Council of Elrond: not the decision, but the summons. The recognition that the problem had grown beyond any single tradition's capacity to manage alone, and that the peoples who would be affected needed to be in the same room. Different languages, different wisdoms, the same table. What Leo XIV published in May 2026 looks like that summons. The encyclical does not resolve what must be done with the instrument. It calls the peoples to gather.

The techno-optimist collapses the tension in one direction. Data will save us, the system will improve, the right guardrails will contain the downside. Keep building. The pure sceptic collapses it in the other. The system is too dangerous, the tools too corruptible, the risk too high. Stop building. Both positions are wrong, because both resolve a tension that has to be lived inside, not resolved. And only the gospel was built for that, for the person who is capable of genuine good and constitutively bent toward using that capacity in service of themselves.

The reformer who knows this can use the tools without worshipping them. Can build without betting the world on the arrangement's permanence. Can hold the coin on its edge, not because they are strong enough, but because they have stopped believing the arrangement will hold it for them.

Karp is a deeply strange man who owns an alarming amount of our data, and he is trying, in his way, to do something genuinely important with it. The warning is not that he is lying about his intentions. It is that sincerity has never been sufficient protection against becoming what you were built to fight.

Leo XIV described researchers and developers who, by working only within their own domain, risk cooperating with the very projects they were built to oppose, perhaps unknowingly. Perhaps unknowingly. The gap between sincerity and unknowing is not a large gap. It is exactly where Karp stands.

The problem of power corrupting is not solved by better power.

It is solved by knowing you are already corrupted.

Human idea ideation. AI assisted writing. Tool: Claude (Anthropic).