On May 25, the leader of 1.4 billion Catholics issued a 43,000-word condemnation of the artificial intelligence industry, warning of an "idolatry of profit" and a new "Tower of Babel." Standing at his side, invited to the unveiling, was a co-founder of one of the largest AI companies on earth. Sections of the document describing the technology's dangers bore the influence of his lab.

The encyclical is the oldest instrument of moral authority in continuous use — a teaching document the Catholic Church has deployed since the second century to address what falls outside the market and above the state. Magnifica humanitas is the first one ever written about a technology. It calls for binding regulation, the protection of children from hypersexualized AI images, and the "disarming" of AI in warfare. On its surface, it is an indictment.

And the company it indicts helped write it.

What an encyclical is for

The design of papal teaching is independence. An encyclical works precisely because it issues from outside — outside Washington, outside Brussels, outside the quarterly earnings call. When Leo XIII wrote Rerum novarum in 1891 on the rights of labor, its weight came from the fact that no industrialist sat on the drafting committee. The Church's leverage over the powerful was that it could not be bought, lobbied, or staffed by the parties it judged. That was the whole point. Moral authority is the one asset that evaporates the moment its source is captured.

Pope Leo XIV chose his name to honor that 1891 encyclical, telling reporters he saw AI as the labor question of his century. The framing positioned the Church exactly where it has always claimed to stand: outside the system, looking in, owing nothing to anyone inside.

The reality of how Magnifica humanitas came to exist tells a different story.

The phases nobody noticed

The relationship did not begin in May. It began the way these things always begin — as outreach so modest that no one thought to track it as strategy.

Each phase seemed like a courtesy. A hackathon is a courtesy. A profile of a friendly priest is a courtesy. A dialogue is a courtesy. None of them looked like the construction of a relationship that would end with an industry's most pointed critic releasing a document the industry had helped shape.

By early 2026 the courtesies had become a campaign. In March, Anthropic met with Christian leaders to ask whether its chatbot could be considered a "child of God." In April, Anthropic, OpenAI, and others convened Hindu, Sikh, and Greek Orthodox leaders to draft principles for "how to infuse models with ethics and morality." Then came the executives at the Vatican gate on April 29. Then the announcement, on May 19, that Anthropic co-founder Christopher Olah would stand with the Pope at the unveiling.

From the Vatican stage, Olah told the audience that AI cannot be stopped and would displace human labor at large scale. The Pope's document, released the same week, contained a passage on AI's unpredictability that observers traced directly to Anthropic's research. The critic and the criticized had arrived at the same warning — because they had drafted it together.

The reversal

An encyclical was designed to be the one verdict an industry could not influence. By May 25, that was no longer true. Not because the Church was corrupted — there is no evidence the Vatican sold anything. Because the industry discovered something more elegant than corruption.

You don't have to capture a moral authority if you helped build it. You only have to be standing next to it when the verdict is read.

the Church spent building an authority that money could not buy
the industry's outreach took to be standing inside it

The mechanism is not a payment. It is participation. When the only institutions with the technical fluency to explain AI's dangers are the labs that build AI, the moral authority that wants to speak credibly about those dangers must source its understanding from them. The Church needed to know what to fear. The companies were the only ones who could explain it. So the explanation of the danger, and the framing of the danger, and eventually the language of the danger, all flowed through the same pipe — from the accused to the judge.

This is why Olah's remarks and the Pope's warnings rhymed. Anthropic has spent two years arguing that AI is powerful enough to be dangerous and that this danger is the reason to take the company seriously. An encyclical warning that AI is powerful enough to threaten human dignity is not a refutation of that argument. It is the most authoritative endorsement of it ever issued. Magnifica humanitas condemns the technology in exactly the terms that make its builders indispensable.

The same move, in a secular suit

On the very day the encyclical was presented, a second moral authority was being profiled — one with no incense and no Latin, but the identical structure.

May 25, 2026
A look at the UK's AI Safety Institute, whose researchers probe AI models for safety gaps, as its work becomes a blueprint for other governments' AI policies
New York Times

The institute is becoming the template that governments around the world copy when they decide how to police AI. It is the secular equivalent of the encyclical — the body that issues the credible, technically informed verdict on whether a model is safe. And it is, per the same report, "staffed by alumni from OpenAI and Google."

The structure is identical to the Vatican's. A body designed to judge AI from the outside is staffed, informed, and shaped from the inside. The Church draws its understanding of the danger from the labs; the safety institute draws its researchers from them. In both cases the institution is real, the work is real, and the independence is a category error — because the only people who understand the technology well enough to govern it are the people who built it.

This is not a religion story or a regulation story. It is one structural pattern wearing two costumes. When a technology becomes too complex for any outside institution to evaluate without insider knowledge, every institution built to evaluate it must recruit from the inside. The verdict stops being external the moment the only available expertise is internal. The Vatican and Whitehall arrived at the same arrangement from opposite ends of the secular-sacred spectrum, and neither had to be corrupted to get there. The structure did the work.

What the document can't say

There is a passage in Magnifica humanitas where the Church apologizes for its own historical slowness to condemn slavery — an institution acknowledging, in the same breath as its AI warning, that moral authorities can take centuries to recognize the harm in front of them. It is the most honest sentence in the document, and it cuts both ways. An institution that needed four hundred years to condemn one system it was entangled with is now issuing a same-week verdict on a system whose dangers were explained to it by that system's manufacturers.

The encyclical asks governments to slow AI down. It may well do real good; the protection of children from synthetic abuse imagery needs no insider to justify it. But the document's deepest assumption — that the Church speaks here from outside the machine — is the one thing the machine has quietly rewired. The referee learned the rules from the players. The verdict is sound. The independence is the fiction.

In 1891, the leverage of Rerum novarum was that no factory owner helped write it. In 2026, the co-founder of an AI company stood at the Pope's side as the verdict against AI was read aloud — and the most damning lines were partly his own. The Church spent two thousand years building the one authority that money could not lobby. The industry didn't lobby it. It joined it.