Is AI a modern Tower of Babel that dehumanizes us? In the face of technological overreach and the tyranny of the average, Leo XIV’s encyclical Magnifica Humanitas proposes a limit to safeguard our dignity. A manifesto for digital disarmament.
In May 2026, Las Vegas hosted for the first time an athletic competition where doping is not a trap to hide, but the very spectacle itself. In the Enhanced Games, swimmers, runners, and weightlifters compete without any anti-doping control, under the argument that the body belongs to the athlete and that the limits imposed by biology are an unfair condition that we must reject. It’s important not to interpret this sad spectacle as an isolated sports scandal. It is a perfectly staged vision of the modern man: the body understood as a system to optimize, performance as the only measure of greatness, and finitude as a factory defect that technology comes to correct.
This way of viewing the human being (as a product to improve) is not born nor exhausted in the sports realm. It is the idiosyncrasy of a significant part of those who today design artificial intelligence systems, people who see progress as the overcoming of homo sapiens and who, without saying it outright, work towards an augmented humanity, hybridized, freed from its own conditions of existence. The doped body and the machine that promises to think for us belong to the same dream. There is a continuity between those who want to swim faster than flesh allows and those who want an artificial intelligence model to write, decide, and discern better than a human could. In both cases, what is excessive is the limit, and what is pursued is a version of man without the rough edges that make him, precisely, human.
I read it not as a practicing Catholic, but as someone who works in technology policy and came across a document that, from a religious framework, names with commendable precision some things that our field of study is still questioning.
On May 25, the Vatican released Magnifica Humanitas, the first encyclical by Leo XIV entirely dedicated to artificial intelligence. The text, like other encyclicals, is an influential religious and political document, starting from the premise that is the opposite of the productivization of the human being. I read it not as a practicing Catholic, but as someone who works in technology policy and found a document that, from a religious framework, names with commendable precision some things that our field of study is still questioning. What follows is an exercise in translation: from the anthropology that the text offers to the governance criteria that interest me. Where the Pope speaks, I mark it with the paragraph number in parentheses and with quotes.
The encyclical opens with two biblical images, the first being Babel. Leo XIV's reading is not that of punishment for pride. Babel is the project of "one language, one technology, one direction" (§7), an effort that, "instead of communion, chooses homogenization" (§7). The sin is not the material fact of the tower but the ideal of uniformity that sustains it: human diversity sacrificed on the altar of a single way of doing things. Later, he gives it a clinical name, the "Babel syndrome" (§10), the pretension of a unique language capable of translating everything, even "the mystery of the person, into data and performance" (§10).
I've been mulling over that idea for some time, without having the exact word, under the formula “the machine of mediocrity,” with a more technical than moral intuition. A language model trained on the corpus of "everything written, captured in images, etc." does not produce ideas: it predicts the most likely continuation to the request or instruction given to it. And, most likely, in such a vast space of possibilities, it is the average. It is statistical recurrence, the center of gravity of what has already been said, a false consensus that presents itself with the (false) certainty of those who know and create. The encyclical describes that mechanism better than I had attempted: the machine learns through "statistical adaptation from data" that "does not imply inner growth" (§99).
The sin is not the material fact of the tower but the ideal of uniformity that sustains it: human diversity sacrificed on the altar of a single way of doing things.
The problem is not mediocrity in itself. There are tasks for which we precisely want an average result. When I ask how to get to a destination, I am not looking for originality; I am looking for the shortest and most traveled route, in the average of possible routes, and that’s perfectly fine. The problem arises when we confuse the average with what is good, with what is correct, or with what is human. Because what is interesting about being human rarely lives at the center of statistics. It lives on the margins, in deviation, in fertile error. Some time ago, I worked on this same idea from another angle (below, the video), that of imperfection as that which connects: the trick that is revealed, the distorted mirror that says something true precisely because it distorts. What moves us on a human level is not the polished but what has a crack through which the possibility of completing with our own imagination slips. A machine trained not to err, to always return the average response, is structurally incapacitated for that.
Homogenization does not stop at the result of an AI model. It intrudes into public discourse. Those who control the platforms and the media have, in the words of the encyclical, a remarkable capacity to "influence the collective imagination and present as desirable a certain vision of reality" (§136). And the most effective mechanism of that power is not open censorship or Pavlovian manipulation but something subtler, what the text calls the "architecture of visibility" (§171): what is amplified and what becomes invisible, what is rewarded and what is penalized, all of that ends up shaping opinions and choosing for us, "generating conformity and self-censorship" (§171). A society mediated by machines that return the average and by architectures that reward conformity tends, by design, towards a single voice. Babel again, but this time building an internal tower.
If the first axis is what is at stake in governing this technology, the second is the question of which vision of the human being is worth fighting for. And here the encyclical says something that must be read carefully, because it runs counter to almost all common sense of our time. "The human being does not flourish despite the limit, but often through the limit" (§118). The phrase inverts the entire operation of the Enhanced Games and the ideology they represent. Where transhumanism reads incapacity, illness, old age, and vulnerability as defects to be corrected, the text re-signifies them as the very space in which a person matures and opens up to others. It is not a pious resignation in the face of suffering but a recognition that finitude constitutes the human instead of mutilating it.
From there it continues to a concept that is the key conceptual element of the document: ontological dignity, that which belongs to each one "by the fact of existing" (§52). It is not earned, not credited, and does not depend on performance. And it is exactly what the logic of optimization does not understand. The encyclical clearly identifies the opposing ideology: that which attributes "greater worth to those who are more efficient and productive" (§51), which turns the person into "a project that must be optimized" (§112) rather than someone called to relate. When efficiency becomes the measure of value, the human being ceases to be an end and becomes a resource, evaluated only by what it yields or produces.
A world organized around records, performance, and the productivity curve already has a written answer about what to do with those who do not perform: exclude them.
What follows in the text is the part that interests me most, because it draws the line connecting philosophical discussions with social justice. If we treat humans as mere perfectible matter, says the encyclical, it becomes easier to consider some people as "less useful, less desirable, less dignified" (§117). The step from the cult of efficiency to the language of "necessary sacrifices" (§117) to achieve it is short, and it is always the same ones who end up paying the price for the supposed optimization of the species. Here anthropology goes from theory to practice: the question of whether the human limit is a defect or a condition of dignity decides who we consider expendable. A world organized around records, performance, and the productivity curve already has a written answer about what to do with those who do not perform: exclude them.
There is a trap in the way we usually discuss these things in the world of technology policy, and the encyclical disarms it in a single sentence. "The first choice is not between a yes or a no to technology" (§9). Binarism (being for or against, being a techno-optimist or a technophobe) is comfortable because it saves the work of thinking and is sterile for the same reason. The text proposes, instead, a word from the ecclesial tradition that turns out to be surprisingly useful for public policy: discernment. Not as private introspection but as collective practice, a "shared discernment" (§6) that draws from sciences, cultures, and experiences, and that the encyclical defines in explicit contrast to the exclusive exercise of power: the Church makes its voice heard, it says, "not to dominate, but to serve communion" (§27).
In this sense, discerning in technopolitical terms is refusing to treat each technological system and its implementations as indivisible blocks that are accepted or rejected entirely. It is asking what they are used for, with what data, under what control, for whose benefit, and with what reparations and responsibilities if something goes wrong. It is distinguishing between the application that helps a doctor read an X-ray and the one that decides who is granted credit without anyone being able to appeal. The same technology almost always presents two distinct moral issues. The yes or no is not enough; discernment is needed to distinguish the cases.
For that discernment to be possible, one must first identify what is being discerned, and here comes the word the Pope chooses for this: "disarm" (§110), in the sense of "disarmament." Disarming AI, he says, is not renouncing it but removing it from the logic of arms competition (military, economic, cognitive) and "breaking this equivalence between technological power and the right to govern." The formulation offered by the encyclical is almost a work program: it is about making AI "discussable, refutable, and therefore livable" (§110). It is not enough to regulate it, it warns; "it is necessary to disarm it" (§110). That disarmament also opens the space to discuss the sense of power and domination that often accompanies technological concepts. It is necessary to “disarm” the meanings. The promoters of oppressive technologies speak in a dialect designed so that certain logics cannot be discussed. Surveillance is called personalization; data extraction, service improvement; control, security. Anyone who wants to seriously intervene in these systems must start by giving things their name back.
Discerning in technopolitical terms is refusing to treat each technological system and its implementations as indivisible blocks that are accepted or rejected entirely.
The method isn't complicated, but it requires work and patience: taking each word from the discursive apparatus, translating it to what it actually designates, and placing back at the center the embodied experience of those affected by technology. The encyclical names the systemic version of that same task when it states that talking about the common good today means "unmasking" the new asymmetry and "naming the new monopolies of AI" (§109). Naming is the first political act.
Here we begin to clarify some keys to inform decisions. The encyclical makes a move I didn't expect from a papal text: it includes among the goods destined for universal use (which belong to humanity and not to their formal owners) "patents, algorithms, digital platforms, technological infrastructures, data" (§67). And it reformulates the old principle of subsidiarity for the digital age. Classical subsidiarity protects individuals, families, and intermediary bodies from being absorbed by the State, proposing that actions and decisions be made at the levels closest to citizens and communities. León XIV observes that in the digital revolution, the level that absorbs is no longer just that of the State: it is the platforms, the major players that "define access conditions, visibility rules, forms of relationship, and even economic opportunities" (§71). From that diagnosis, a concrete agenda is deduced, one of: "independent audits, transparency in algorithms, equitable access to data" (§71), advocating for concrete pathways for exercising rights.
Regarding data, the encyclical is more radical than much of the dominant regulatory debate. It argues that its ownership "cannot be entrusted solely to the private sector" (§108), because they are the result of contributions from many and must be managed as a common good. And it concludes with a formulation that offers a clear guide for creating AI models and other technical tools: social justice is not something that is safeguarded after technology is adopted, but rather "a condition that must be put into practice from its design" (§109).
It offers a clear guide for creating AI models and other technical tools: social justice is not something that is safeguarded after technology is adopted, but rather a condition that must be put into practice from its design.
This connects with the idea of technological sovereignty: the autonomy of a person, a community, or a country to establish its own rules and ensure its freedom in the design, use, implementation, and relationship with technology. This has dimensions that range from political, jurisdictional, and economic to personal and community dimensions related to psychic autonomy. For countries that do not have their own AI developments (the majority of the globe), the issue is not trivial. Depending entirely on models trained elsewhere, with different values, on different data, is a new and elegant form of domination. The answer is not isolation but a deliberate policy: open source wherever possible, specific models instead of the fantasy of a one-size-fits-all model that knows everything, strict but progressive regulations tailored to specific purposes instead of prohibitions that only the powerful can circumvent. A fairer AI is actively built, without asking for permission.
There remains an objection that deserves to be taken seriously, because it comes from people I respect. When an institution like the Church sits down to dialogue with those who build these systems, some see it as a capitulation: the risk of lending moral legitimacy to the very companies whose power the document denounces, of playing into the hands of the powerful instead of listening to those whom technology harms. It is an honest critique and has a true foundation. But it rests, it seems to me, on a narrow idea of how to change the world.
It is necessary to distinguish between strategy and tactics. Strategy is the calculation of someone who has their own place, a position from which to organize the field and negotiate on equal terms with other powers. Tactics is the art of those who have no place, of the weak who operate within the other's terrain, taking advantage of cracks and the opportune moment. They are two legitimate ways of acting on an order that exceeds one, and their value depends on where the actor stands. The Church, which has been doing real politics for centuries in front of kings, empires, and markets, plays a strategy game: it has a place from which to speak to the powerful without needing to legitimize itself. The civil society that pressures from the outside, that denounces, that organizes those affected, plays a tactical game, and does so because that is where it can. The encyclical defends the former in all its letters when it claims a realism that "does not renounce changing the world" (§218) and starts by clearly seeing the interests and power relations to calculate what can be achieved, and when it reminds us that the vocation of diplomacy is to dialogue even with the "most uncomfortable interlocutors" (§224), with those who are not even considered legitimate for negotiation.
The problem with artificial intelligence systems is not resolved by choosing between confrontation and conversation; it is resolved by doing both at the same time, from different positions, by different people, who recognize themselves as allies and not rivals.
The game from the inside and the game from the outside do not cancel each other out: they are needed. Those who pressure from the street make credible those who negotiate at the table, and those who negotiate at the table open doors that the street alone cannot break down. The problem with artificial intelligence systems is not resolved by choosing between confrontation and conversation; it is resolved by doing both at the same time, from different positions, by different people, who recognize themselves as allies and not rivals. That a millennia-old institution has decided to step into the territory of those who build the future, and has brought to that table a document that defends the dignity of the discarded, the common destiny of data, and justice from the design, is not a cost of negotiation: it is the result of taking seriously that to change something, one must first be willing to talk to those who have the power to change it. The hard part comes later, in the implementation: in whether those principles become functioning legislations, audits that have institutional strength, regulations that compel, models that are trained and deployed differently. That will be seen. But the table, for now, is set, and having succeeded in getting the Church to sit down to converse on the terms of the encyclical Magnifica Humanitas is undoubtedly an achievement that inspires our objectives, tactics, and values.
