The Pope just called your AGI roadmap a heresy

What happened

On May 28, Pope Leo XIV — the former Cardinal Robert Prevost, elected last year as the first American pope — published his inaugural encyclical, *Dilexi Te* ("I Have Loved You"). Buried under the expected pastoral themes is a passage that has landed harder in San Francisco than in Rome: a direct condemnation of what he calls technological messianism, the belief that artificial intelligence and biotechnology will deliver humanity from suffering, mortality, or its own moral limits.

The encyclical does not name OpenAI, Anthropic, or Google, but it does not need to — the rhetorical targets are unmistakable to anyone who has read a Sam Altman blog post or a Marc Andreessen manifesto. Leo singles out the language of "superintelligence," the framing of AGI as inevitable, and the transhumanist project of "merging" with machines as a contemporary form of idolatry: building a golden calf out of GPUs. He draws an explicit line back to *Laudato Si'*, Francis's 2015 environmental encyclical, but where Francis warned about a "technocratic paradigm," Leo escalates to theological language — *messianism*, *salvation*, *false eschatology*.

The Economist's read, echoed in the 202-upvote Hacker News thread, is that this is the first papal document to treat the AI industry's own self-mythology as a doctrinal adversary rather than a policy problem. Leo, who taught mathematics before entering the Augustinian order, is not technologically illiterate. The text shows familiarity with foundation-model capabilities, the alignment debate, and the labor displacement literature. It is not a luddite document. It is a hostile one.

Why it matters

The instinct in our industry is to file this under "old institution complains about new technology" and move on. That instinct is wrong, and the reason is operational, not spiritual.

The Catholic Church is the largest non-state operator of hospitals, universities, and primary schools on earth — roughly 5,500 hospitals, 1,800 universities, and 220,000 schools across 195 countries. Every one of those institutions has a procurement process. Every one of those procurement processes is, downstream, governed by the moral teaching of the Holy See. When the Vatican issues an encyclical naming a specific technological posture as incompatible with Catholic social teaching, it does not stay in seminaries. It shows up, eventually, in a Notre Dame IT contract, in a Bambino Gesù pediatric-AI vendor questionnaire, in a Loyola Marymount research-ethics review.

The more interesting second-order effect is regulatory. The EU AI Act's text on "human oversight" and "fundamental rights impact assessments" was lobbied for by, among others, the Pontifical Academy for Life, which co-signed the 2020 *Rome Call for AI Ethics* with IBM and Microsoft. Leo has now given European regulators a theological superstructure to cite when they want to push back on American AI lobbying — and "the Pope says no" is a stronger card in Brussels, Dublin, and Warsaw than American executives tend to remember. The Rome Call framework — transparency, inclusion, responsibility, impartiality, reliability, security and privacy — was already showing up in procurement language. *Dilexi Te* sharpens its teeth.

The community reaction on Hacker News split predictably along two axes. The first camp dismissed it as irrelevant — religion has no jurisdiction over technical roadmaps. The second, and more interesting, camp pointed out that the Church has been the most durable institutional skeptic of human-enhancement discourse since the bioethics debates of the 1990s, and has been consistently right about which way those debates would eventually cut (genetic enhancement, embryo selection, gain-of-function research). The pattern is: the Vatican stakes out an unfashionable position twenty years before the median bioethics consensus catches up. Betting against the Church on questions of "what should we not build" has historically been a worse trade than the AI industry assumes.

There is also the labor angle, which the secular coverage has under-weighted. *Dilexi Te* spends more text on AI's effect on dignified work than on AGI eschatology. Leo, like Francis, treats unemployment not as a market outcome but as a violation of human dignity, and frames productivity gains that flow to capital while displacing workers as a structural sin. This is not new Catholic teaching — it goes back to *Rerum Novarum* in 1891 — but applying it explicitly to model-driven automation gives unions, works councils, and Catholic-affiliated worker centers a moral vocabulary they did not have last week.

What this means for your stack

If you build for enterprise, the practical change is procurement language. Expect the next round of RFPs from Catholic-affiliated hospitals, universities, and NGOs to include questions about "human oversight in clinical decision support," "transparency of training data," and explicit prohibitions on autonomous lethal applications — not as boilerplate, but as scoring criteria. If your product cannot answer those questions cleanly, you will lose deals you used to win on demo quality alone. The Rome Call signatories — IBM, Microsoft, Cisco — already have compliance teams set up for this. Smaller vendors do not.

If you build consumer AI, the second-order effect is reputational cover for skeptics. Until this week, an executive who wanted to slow-walk an AI rollout had to frame it in terms of "responsible deployment" — a phrase Silicon Valley has successfully neutralized. They now have a 30,000-word theological document to cite, which is significantly harder to spin as anti-innovation reactionism. Watch for European boards, in particular, to cite *Dilexi Te* in declining specific use cases.

If you build infrastructure — the foundation-model labs themselves — the implication is harder. The encyclical does not call for moratoria or bans. It calls for a posture: humility about what these systems are, rejection of the framing that they are humanity's successor, refusal to participate in the narrative that AGI is inevitable and therefore moral debate is futile. That posture is incompatible with current frontier-lab marketing. The labs will have to decide whether to engage substantively or wait for the news cycle to pass. Anthropic, given its explicit interpretability-and-safety framing, has the easiest path. OpenAI, whose CEO has publicly mused about AGI as a quasi-religious event, has the hardest.

Looking ahead

The Vatican plays a long game measured in decades, not quarters, and its track record on technology questions is better than the industry's vanity allows. *Dilexi Te* will not change a single model release schedule this year. But five years from now, when an EU regulator drafts the successor to the AI Act, or a hospital system writes an autonomous-diagnosis policy, or a university negotiates a partnership with a foundation-model lab, the language they reach for will have been shaped by this document. The pope did not ban AGI. He gave the institutions that distrust it a vocabulary — and in regulated markets, vocabulary is leverage.