White papers arrive as PDFs. Frameworks arrive as press releases. Product updates arrive as changelogs, usually wearing the expression of someone who has already decided the future is inevitable and would appreciate it if everyone else stopped asking where the invoices go. Moral infrastructure arrives differently. Sometimes it arrives as thunder.
On 15 May 2026, deliberately marking the anniversary of Rerum Novarum, Pope Leo XIV signed Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence. Ten days later the document entered the public room with the full ceremonial weight of an institution that thinks in centuries and still knows how to use a headline. That timing is not decoration. Rerum Novarum was the Church's great intervention into the industrial question: labor, capital, the factory, the worker, the moral disorder of treating human beings as disposable inputs to production. One hundred and thirty-five years later, the new thing is no longer the factory floor. It is the machine that speaks, summarizes, ranks, predicts, persuades, and increasingly stands between the human person and the institution that claims to serve them.
The encyclical's most important word is not "intelligence," and it is not even "humanity." The word is disarmed. That word matters because it refuses the usual little dodge. Regulation can still imagine the weapon as legitimate if the paperwork is tidy. Alignment can still imagine the system as acceptable if the output is smooth enough, polite enough, useful enough, and visibly compliant with the local policy stack. Ethics can become an ornamental layer added after the architecture has already coupled itself to speed, scale, surveillance, profit, labor extraction, military appetite, and institutional convenience.
Disarmament starts earlier. It asks what the system has been armed to serve, not whether the model can answer or whether the interface is friendly or whether the dashboard has a responsible AI page with blue icons and a paragraph about human dignity. The question is older and harder: what pattern of power does this system extend?
That is the force of the encyclical. The document is a claim of jurisdiction over the moral imagination of AI before technical categories harden into governance defaults. It is not a technical policy document, and it does not aspire to be one. The Pope is not merely warning that artificial intelligence may be misused. Everyone says that, and misuse has become the beige cardigan of AI ethics: respectable, soft, and somehow always available when the actual structure of power needs to leave the room early.
The encyclical makes a stronger move. It argues that AI cannot be understood only as a tool whose moral status begins at the moment of use. A tool built inside a political economy already carries a direction. A model trained, deployed, priced, distributed, integrated, and procured under particular incentives does not enter the world naked. It enters with a wardrobe, with patrons, with default use cases, with a theory of what counts as value.
That is why "disarmament" is the word that changes the frame. The weapon is not only the drone, the targeting system, the surveillance platform, or the predictive policing stack. The weapon is also the logic that turns every human relation into an optimization surface. It is the invisible settlement that lets speed outrank judgment and consent, and that lets prediction overwrite testimony before anyone notices the substitution.
The encyclical enters that settlement and says: no. No, the human person is not reducible to data. No, moral agency does not emerge because a system produces fluent moral language. No, convenience is not enough. No, alignment with "human values" cannot mean alignment with the values of whoever owns the training run, the platform, the cloud, the procurement channel, or the classroom dashboard.
The useful irritation
There is a useful irritation here, and the irritation is part of the gift. The Church is not a neutral observer. No institution that has survived two millennia gets to pretend it has never confused moral authority with control. The encyclical itself reportedly contains an apology for the Church's complicity in slavery, which is a pressure mark in the marble rather than a side note. It reminds the reader that institutions can sanctify harm while believing they are defending truth, that they can lag centuries behind the dignity they later claim to have recognized all along, and that they can be necessary witnesses and dangerous custodians at the same time.
That does not weaken the document. It makes the document more serious. A Church that admits it has misread the human person before has no cheap route to certainty now. Its warning against invisible moral infrastructure lands with extra weight precisely because the warning also curls back toward itself. Who builds the invisible moral infrastructure? The model lab, the state, the market, the school, the doctrine, the family, the procurement office, the safety team, the user interface, the classroom, the priest, the engineer, the person who decides which voices count as stakeholders and which arrive only as edge cases. The encyclical does not escape that problem. It stages it.
The jurisdictional reading
That is why the most interesting reading is neither pious agreement nor theatrical denunciation. The useful reading is jurisdictional. The Pope is claiming that the human boundary cannot be left to product teams, markets, militaries, or safety benchmarks alone. He is asserting that the question "what is the human?" belongs upstream of deployment, and he is asserting, as every institution does when it speaks at this scale, that his tradition has standing to define the boundary. That is the live wire.
AI governance keeps presenting itself as if the main dispute is technical: improve the model, reduce hallucination, document the data, harden the policy layer, audit the outcomes. Those things are real and necessary. They are not enough. Beneath them sits the jurisdictional question: who gets to name the moral perimeter before the machine is normalized?
The encyclical answers that the human person comes first, and the Church has something to say about what that means. The market answers differently — the user comes first, especially when the user is convertible into subscription revenue, engagement, data exhaust, or enterprise lock-in. The state answers differently again, with public safety, national security, administrative efficiency, emergency capability, sovereign advantage. The lab answers with another vocabulary entirely: capability, alignment, interpretability, evals, red lines, responsible deployment. The person at the edge of the system often answers more simply: will this thing help me, harm me, replace me, misread me, expose me, route me, price me, profile me, or make it harder for another human being to hear what I meant?
Monday begins there, with competing claims of authority rather than a settled answer. The encyclical is thunder because it insists that AI is more than a tool problem. It is a formation problem, a moral imagination problem, a jurisdiction problem. The week ahead asks whether other forms of jurisdiction can draw the line more practically, more locally, more enforceably, or more honestly. Because doctrine can name domination, and naming is not yet disarmament. A line spoken from a pulpit still has to survive the workshop, the contract, the courtroom, the category fight, and the room where the machine explains itself back to us in the voice we most wanted to hear.