Pope Leo XIV Identified the Right Problem
In my last post I argued that the AI debate runs simultaneously at five levels (task, person, organization, economy, civilization) and that no one has agreed on which level should have standing. The civilization level, where the hardest questions live and where costs tend to concentrate, keeps getting treated as decorative. Nobody is lying. Everyone is describing what they can see from where they are standing. What’s missing is the map.
Pope Leo XIV’s May 2026 encyclical, Magnifica Humanitas, is the most serious attempt in recent memory to argue from that level. This is what that looks like.
I work in institutional research at a Lasallian university. My job is to measure things: enrollment, retention, outcomes, the signals that tell us whether students are getting what they came for. I am also a practicing Catholic. For most of the last three years I have watched the AI conversation happen around higher education from both of those angles, the data side and the formation side. My conclusion is that the governance conversation has been asking the wrong question. The governance conversation has been running the same four tracks for years: alignment, safety, governance, labor displacement. These are real problems. But they share an unexamined premise, that the central challenge is managing a technology we have already decided to deploy as widely as possible. Every major framework of the last three years, from the EU AI Act to major university AI policies, reflects this. They are engineering questions dressed in policy language.
Pope Leo XIV does not accept that premise.
He opens with an anthropological claim, not a regulatory one. Artificial intelligence, he writes, does not undergo experiences. It does not possess a body. It does not feel joy or pain. Its “learning” is statistical adaptation, not inner growth. This is not a statement about technical limitations. It is a statement about what learning is: a process inseparable from having something at stake, from the capacity to be changed by what you encounter, from the development of a self that persists through time and carries what it has known. Statistical adaptation produces outputs. It does not produce persons.
This is not just a philosophical claim. Research from MIT’s Media Lab measured what happens neurologically when students write with AI assistance compared to writing without it (Kosmyna et al., 2025). The findings were not subtle. Students who relied on AI showed weaker neural connectivity, collapsed recall, and low ownership of essays they had produced minutes earlier. The essays scored well. The students who wrote them could not quote from them. Output and formation had come apart. This is exactly where person-level cost hides in the five-level framework: invisible in output metrics, accumulating later, in ways no dashboard captures. The encyclical names the mechanism. The neuroscience supports the consequence.
I see a version of this in the data I work with. The signals that show up in retention and advising, students performing adequately on measurable outcomes who are somehow not developing, are not new. But they are sharper now. The encyclical gave me language for something I had been watching without words for it.
The second thing Pope Leo gets right is about power. He argues that the concentration of digital infrastructure outside state or democratic oversight is a new form of domination. Income from capital, the encyclical notes, risks replacing income from labor in ways that echo the crisis that produced Rerum Novarum in 1891. This is not a prediction. It is a diagnosis of what is already underway. Distributional economists have documented that AI is capital-biased: each doubling of AI innovation corresponds to a measurable decline in labor’s share of national income (Acemoglu & Restrepo, 2022; Minniti et al., 2025). The pope and the economists are looking at the same thing.
The third is about education, and it is where the argument lands hardest. The encyclical calls for an educational alliance for the digital age centered on formation rather than competency certification. Universities have spent two decades building assessment infrastructure to demonstrate that students have acquired measurable skills. That apparatus cannot capture what education is actually for. Formation requires cognitive friction, relationship, and accountability. It requires students to encounter difficulty without immediate rescue.
The Lasallian tradition has been making this argument for three hundred years. De La Salle did not build schools to certify competencies. He built them because he believed that the poor deserved an education that developed them as full human beings, one that required something of them and gave something back that could not be put on a rubric. I work at an institution that carries that tradition. Watching it navigate the AI moment, I find myself asking daily whether we are actually living it or just citing it.
The encyclical does not answer every question. It offers no alignment strategy, says nothing about compute governance or model interpretability, and provides moral orientation rather than technical implementation. Anyone looking for a papal AI policy platform will come away disappointed.
But that is not the right test. The question is what you are trying to preserve when you build those frameworks. That is the prior question, the civilization-level question the governance conversation keeps deferring. That is what the encyclical answers.
The right test is whether it identified the correct problem. On that test, Pope Leo has done something most of the AI governance conversation has not. He has asked what human development requires, what education is for, and what kind of institutional life makes those things possible. You cannot build a safety framework without knowing what you are trying to protect. You cannot govern AI’s role in education without knowing what education is supposed to produce in a human being.
Most institutions are not asking that question with any seriousness. Some of us work inside institutions that were founded precisely to ask it. That is either an advantage or an indictment, depending on what we do next. The map exists. The question is whether we use it.
References
Acemoglu, D., & Restrepo, P. (2022). Tasks, automation, and the rise in US wage inequality. Econometrica, 90(5), 1973–2016. https://doi.org/10.3982/ECTA19815
Minniti, A., Prettner, K., & Venturini, F. (2025). AI innovation and the labor share in European regions. European Economic Review, 177, 105043. https://doi.org/10.1016/j.euroecorev.2025.105043
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2506.08872
Leo XIV. (2026). Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence[Encyclical letter]. Libreria Editrice Vaticana. https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html
Urmeneta, M. (2026, May). Nobody is lying about AI. https://allthingy.com/nobody-is-lying-about-ai/
