Pope Leo XIV has called for artificial intelligence to be “disarmed.” This is, at first glance, the sort of papal phrase that invites both applause and eye-rolling. It sounds noble. It also sounds vague enough to be filed beside every other high-minded institutional plea to make technology “serve humanity,” as if the technology were a disobedient but promising intern.
But the phrase deserves more respect than that.
Source note: This essay responds to Pope Leo XIV’s first encyclical, Magnifica Humanitas, signed 15 May 2026 and presented by the Holy See on 25 May 2026. The encyclical addresses “safeguarding the human person in the time of Artificial Intelligence” and calls for AI to be “disarmed,” meaning freed from logics of domination, exclusion, war, and death. See the full Vatican text of Magnifica Humanitas; Pope Leo XIV’s 25 May 2026 address presenting the encyclical; Vatican News, “Pope Leo presents Magnifica Humanitas calling for disarmament of AI”; and Christopher White, “Pope Leo calls to ‘disarm’ AI in major document, warns of technologic threats to humanity,” National Catholic Reporter, 25 May 2026. The “babies strapped to car bumpers” analogy is credited to Bruce, in conversation at a late-1970s Dungeons & Dragons table, and is used here as a moral thought experiment about deterrence, not as a policy proposal.
To “disarm” AI should not mean to abolish it, fear it, or pretend that software has become a demon crouched in the server rack. It should mean stripping AI of the logic of domination: the assumption that because a thing can be automated, scaled, optimized, surveilled, or weaponized, therefore it may be. That is the ancient barbarism of power dressed in new syntax.
The immediate issue is war. Pope Leo is right to reject the delegation of lethal or irreversible decisions to artificial systems. There must be a human chain of responsibility. Not a ceremonial human. Not a tired officer rubber-stamping machine recommendations at impossible speed. Not a procurement committee hiding behind a vendor’s confidence score. A real human responsibility, with names, offices, laws, consequences, and blame.
Otherwise we will have built the perfect modern atrocity machine: everyone participated, no one decided, and the dead were merely outputs.
The familiar reply arrives at once: our enemies will build these systems, so we must build them too.
This argument is sometimes presented as realism. Often it is merely surrender wearing a uniform.
“The baddies have it, so should we” is one of the most dangerous lines of reasoning in public life. It accepts the enemy’s moral architecture before the first shot is fired. If kleptocracies, prison-states, and lunatic regimes organize the world around coercion, surveillance, and expendable human beings, then the answer is not to become a better-managed version of them. The answer is to make their model economically, culturally, and politically unsustainable.
That means isolation where isolation is warranted. It means reducing dependency on regimes that use trade as leverage and technology as a leash. It means defending liberties at home instead of sacrificing them in the name of competition abroad. It means building societies so visibly freer, fairer, more resilient, and more humane that the authoritarian bargain looks as shabby as it is: obedience in exchange for managed decay.
And it means education. Not merely technical education. Not just STEM pipelines feeding the next procurement cycle. Humanistic education. Historical memory. Moral reasoning. Civic courage. The study of language, law, religion, philosophy, art, propaganda, empire, and failure. People trained only to build systems are easily recruited by systems. People trained to ask what a system is for are harder to conscript.
The counterargument is deterrence. Sometimes bad actors do not stop because they are morally embarrassed. They stop because the cost of aggression is made unbearable. This argument cannot be dismissed casually. Defensive strength matters. Open societies that cannot defend themselves become museums, then provinces, then footnotes.
But deterrence is often mistaken for a solution when it is really a suspended catastrophe.
Nuclear weapons are the classic case. We are told they have preserved great-power peace since 1945. Perhaps. But that is not the same as proving that nuclear deterrence is morally coherent or permanently stable. It may only prove that we have survived the first several spins of the cylinder.
A friend of mine once proposed a grotesque thought experiment at a Dungeons & Dragons table in the late 1970s: what if we strapped babies to the bumpers of all cars? Accident rates would presumably collapse. Drivers would become exquisitely careful. Tailgating would vanish. Speeding would become unthinkable.
Would that make it a good policy?
Of course not. It would be monstrous. The fact that terror can regulate behavior does not redeem the terror. It proves only that human beings can be made careful by holding innocents hostage.
That is the moral structure of nuclear deterrence. Peace is maintained by the standing threat of mass civilian annihilation. We call this stability because the alternative is too frightening to name. But it is not peace in any mature ethical sense. It is hostage management.
The same danger now shadows AI.
If autonomous weapons become normal because adversaries may use them, if mass surveillance becomes acceptable because hostile states already do it, if algorithmic manipulation becomes necessary because propaganda networks are faster than parliaments, then the free world has already conceded the central point. It has agreed that human dignity is conditional, that liberty is negotiable, and that our institutions exist only until fear drafts a better memo.
That is not defense. That is mimetic corruption.
A serious doctrine for AI and security would reject both naïveté and imitation. It would say: yes, we need defensive capacity; no, we do not grant machines authority to kill. Yes, we need intelligence tools; no, we do not normalize total surveillance. Yes, we need industrial resilience; no, we do not build an economy in which human beings are merely data exhaust, content moderators, warehouse bodies, and retraining liabilities.
The point is not to leave democratic societies helpless. The point is to defend them without hollowing them out.
AI should be governed by several hard principles.
First, no artificial system should make final lethal decisions. Machines may assist targeting, detection, logistics, translation, and defense. They should not be permitted to decide who dies.
Second, every consequential AI system must preserve a chain of responsibility. If an algorithm affects liberty, employment, credit, medical access, education, welfare, policing, migration, or military force, then someone must be answerable for its design and use.
Third, economic concentration is a security risk. When a handful of firms control the data, models, infrastructure, compute, talent, and lobbying power behind AI, the public interest becomes a licensing problem. Civic life should not be downstream from a server farm owned by oligarchs.
Fourth, democratic states should pursue technological non-dependency. That means export controls on repression tools, sanctions against firms enabling authoritarian surveillance, domestic capacity for critical infrastructure, and alliances among free societies that treat human rights as operating conditions, not decorative language.
Fifth, education must become a strategic instrument again. A society that cannot teach its citizens how to detect manipulation, weigh evidence, understand history, and defend liberty will lose to its own devices before it loses to foreign enemies.
Pope Leo’s language may be theological, but the institutional problem is plain. Technology is never just a tool. It carries the values of those who fund it, design it, regulate it, and deploy it. A hammer is simple enough. AI is not a hammer. AI is a system for classifying reality and acting upon it at scale. That makes it political before it is technical.
The great error is to think that ethics is a brake placed on innovation after the machine has already been built. Ethics is steering. Without it, speed is only a faster way to hit the wall.
So yes, disarm AI. Not by smashing the machines. Not by pretending that hostile regimes will be moved by our purity. Not by confusing caution with weakness.
Disarm AI by refusing to let it inherit the logic of empire, oligarchy, and mechanized death.
Disarm it by denying it the right to become judge, jailer, executioner, employer, teacher, priest, and sovereign.
Disarm it by building societies that do not need to become monstrous in order to survive monsters.
And above all, disarm the argument that whatever frightens us may therefore rule us.
No comments:
Post a Comment