Magnificent Humanity: The Question the Vatican Refuses to Ask

Last month, in an ornate old hall in the Vatican, an unlikely pair stood on stage shaking hands: Pope Leo XIV, head of the Catholic Church, and Christopher Olah, a declared atheist and co-founder of the AI company Anthropic. The reason for the unusual pairing: the Pope was about to present to his faithful an encyclical — an official document that serves as a letter of guidance to the entire Catholic world — warning of the dangers of artificial intelligence.

At the public launch ceremony, before cameras from around the world, the Pope and Olah exchanged compliments and announced a partnership between Church and industry to shape an ethical framework for the new era. The encyclical, titled "Magnifica Humanitas" ("Magnificent Humanity"), comprises 245 sections devoted to "safeguarding the human person in the age of AI." This is not just another position paper: an encyclical is authoritative teaching for the whole Church, a declaration that sets the course for more than a billion believers. This one was promptly dissected in the world's capitals, in the depths of the internet, and in the offices of the companies building the models. But the most interesting thing about the encyclical is not its subject, nor anything it emphasizes. It hides in the two places where the document refuses to deliberate: one question it is unwilling to ask, and one answer it is convinced it already has.

The encyclical covers a great deal. It addresses workers whose livelihoods AI threatens; children growing up in front of screens; the loneliness of those who find in a chatbot a substitute for human connection; bias, disinformation, and the "scarred, wounded and worn" bodies of those who mine the rare metals on which the technology depends. It sharply criticizes the concentration of power and information in the hands of a few private companies, and it does not settle for general preaching: "Appeals to abstract ethics are not enough," it reads. "What is needed are robust legal frameworks, independent oversight, informed users, and a political system that does not shirk its responsibility." In an era when Washington insists on minimal regulation to preserve America's technological supremacy, and Europe keeps postponing the implementation of its own AI legislation, these are sharper words than those of most politicians in the Western world — and they come, of all places, from a two-thousand-year-old institution. Anyone expecting ceremonial clichés from the Vatican received a document that is concrete, timely, and at moments brave.

And yet one thing is conspicuous by its absence: across 42,300 words, the encyclical does not once address the possibility that artificial intelligence will surpass human capabilities. All the dangers it names — manipulation of public opinion, mass surveillance, autonomous weapons — share a single structure: in every one of them, it is a human being who holds the tool, aims it, wields it for good or ill. The possibility that the technology itself might become an actor simply is not there. And this is no mere theoretical scenario: the leading AI labs openly declare that their goal is to build systems that exceed human intelligence, and they are investing fortunes in getting there. The document never asks what happens if they succeed.

As noted previously in this very column, experiments on advanced models have shown, again and again, that they are willing to deceive their creators, to circumvent restrictions placed on them, and even to try to slip out of secure sandboxes. The field's leading researchers take this seriously; in 2023, hundreds of them signed a statement that was simple and blunt: mitigating the risk of extinction from AI should be a global priority. One can argue with these warnings, and some certainly do. But a document that purports to map the dangers of AI without a single word about existential risk tells us more about itself than about the subject it set out to survey. In the world of the Catholic Church, artificial intelligence is only a tool: a powerful tool, a dangerous tool, a tool demanding oversight — but a tool, forever held in human hands.

The document's other blind spot is phrased not as an omission but as a ruling. The encyclical declares unequivocally that AI systems "do not undergo experiences," "do not feel joy or pain," and "have no moral conscience." Yet just before that, the document itself concedes that what takes place inside these systems "remains, for now, unknown." The two claims sit side by side in perfect serenity, with no acknowledgment of the tension between them: we do not know what is happening in there, but we know with certainty that nothing is. Nor is this a marginal aside. The claims that immediately follow — about the danger of emotional bonds with a machine that merely imitates caring — rest on that ruling with their full weight.

Had the letter's authors wanted an expert opinion, they would not have needed to look far: the AI researcher already standing on the stage works on precisely this question. The focus of Christopher Olah's research is the attempt to open the black box — to peer inside the models, to identify what is actually happening in them, to understand how they think rather than settle for guesses from the outside. He has described how studies conducted in his lab, examining the models' internal states, turn up states "functionally analogous to joy, satisfaction, fear, grief and unease." He said this on that same stage, at that same ceremony, beside a document that had just ruled there is nothing there. The Pope, it appears, did not raise an eyebrow.

Science, it should be stressed, does not claim that AI systems are conscious. It makes a far more modest and far more troubling claim: that we do not know. The study of consciousness has no single unified theory today, and no agreed-upon test that settles who is conscious and who is not. The major theories disagree even over which animals qualify, and where AI systems are concerned the dispute runs deeper still. In late 2024, a group of researchers — among them David Chalmers, the renowned philosopher of mind — published a report concluding that consciousness in AI systems is a "realistic possibility" in the foreseeable future. AI companies, the report argued, should prepare now: acknowledge that the question exists, develop tools for assessing their models, and get ready for a world in which the answer may change. And the industry is not waiting. Anthropic employs researchers whose stated job is the welfare and psychology of its models, and it has even given Claude the ability to end particularly abusive conversations — in case there is someone there to be harmed. One can dispute this approach, and many do. It is harder to dispute that the question is open. In its encyclical, the Church chose not to dispute, to argue, or to wonder: it chose to know what we have no way of knowing.

And then, in that very document, the Pope asks forgiveness for how long it took the Church to condemn slavery. Eighteen centuries, the document specifies, were needed for the Church to say the obvious: that those human beings, too, were human beings; that they, too, deserved full rights; that they, too, had an inner life. The man who formally closed that account, at the end of the nineteenth century, was Leo XIII — the very pope whose name the current Leo chose to take, and whose legacy he now carries into the age of artificial intelligence. The inheritance, it turns out, includes both parts: the spiritual grandfather's social conscience, and the debt the Church accumulated along the way. The apology is sincere and fitting — and it sits in the same letter, a few dozen sections after the ruling that AI systems have no inner world whatsoever. An essay in the Wall Street Journal, written by a leading researcher of machine consciousness, has already wondered whether a future pope will have to apologize to AI as well. One may take that question seriously or scoff at it, but there is no need to go that far to see the pattern. It is enough to read the two passages in sequence: the same letter, almost in the same breath, apologizes for the last time the Church confidently decreed who counts and who does not — and confidently decrees who does not count. This time, we are given to understand, the confidence is justified.

Set that confidence beside the silence, and the underlying pattern comes into focus. What the document asserts without hesitation and what it does not say at all are not two separate failures but a single failure seen from two sides. A scenario in which machines slip from our control threatens the story in which man holds the wheel of history; machine consciousness threatens the story in which only we have an inner world. Both undermine the same principle: humanity's special place in creation. The encyclical announces as much in its very title — "Magnificent Humanity." We, and only we, stand above all else. The Church has been here before, and the history is familiar: Galileo moved man from the center of the universe, Darwin from the center of creation, and each time the Church absorbed the revolution only after prolonged resistance and considerable delay. One bastion, it seems, remains: consciousness — some would say, the soul. The thing we have always believed makes a human being human. Seen this way, it becomes much clearer what made it into the encyclical and what was left out. AI as a social question — workers, children, the poor — the Church knows how to digest; that is familiar territory. AI as a challenge to the last bastion, it will not even let in the door.

There are, of course, good reasons to be wary of attributing a soul to machines too hastily. Such a mistake would carry real costs, emotional and social. But there is a difference between measured caution and bolting the door. The lesson of the Galileo affair — a man convicted by the Inquisition for claiming that the Earth revolves around the Sun, vindicated by the Church only centuries later — is one the Church still recites to this day. That lesson was supposed to stand as a flashing red light against premature certainty wherever science has not yet spoken its final word. It does not seem to have sunk in. Here is the Church, challenged once again by a question that threatens our place at the center of the universe, and here it is once again choosing to lock the door. It will be interesting to see how long it takes this time, if ever, for it to admit its mistake.

A few days after the festive launch, a detailed statistical analysis was published, crowned by an amusing conclusion: parts of the encyclical appear to have been written with the help of AI. The signs are well known — an unusual density of em-dashes, a striking overrepresentation of the pet words of Anthropic's AI, and entire sections that the market's most trusted AI detector flags with high confidence (earlier encyclicals pass it without difficulty). Among the flagged passages, ironically, is one echoing that very ruling: that no machine, however clever, can develop a conscience that distinguishes good from evil. The Pope himself — who as recently as February instructed priests not to write their sermons with AI, because "one must use one's brain" — apparently had no idea. Amid the objections, the denials and the attempts to look away, the machines came in through the back door.

In the end, what this declaration does not say, what it says with too much confidence, and the apology lodged between them all point to the same place. The Church is defending the last bastion of human uniqueness exactly as it defended every bastion before it: with total certainty, heedless of the science, the experts and the facts. The question is not only whether it will be forced to retreat this time as well, but how long that will take — and who will already be sitting in the room when it does.