- AI and religion now broadly agree that today's machines are not conscious.
- Pope Leo XIV's encyclical treats science as a partner, not a foe.
- Its deepest worry is concentrated power over AI, not the technology itself.
Most popes hand the job to a cardinal, but Pope Leo XIV walked out and presented his first encyclical himself.
The room happened to include several of the people building the very technology he had written about. Among them was Chris Olah, the co-founder of Anthropic, who was invited to speak at the occasion.
The document is Magnifica Humanitas, signed on 15 May 2026 and released ten days later by the Vatican. It is the Catholic Church's first encyclical built around artificial intelligence, its fullest word yet on AI and religion, and the occasion was chosen with some care.
He signed it on the 135th anniversary of Rerum Novarum, Leo XIII's 1891 letter on factory labor. The Pope might see that AI is to this century what the steam-driven factory was to the last, and that Rome intends to have views about it.
Pope Leo XIV signs the Mignifica Humanitas. Photo credit: Vatican Press.
The headlines could reach for the obvious story, a pope sounding the alarm about the machines. Read the text itself, though, and the Pope is reluctant to treat science as the enemy. Which probably should not come as a surprise, as Pope Leo XIV holds a Bachelor of Science in mathematics.
Technology Is Never Neutral, the Pope Argues
Leo's starting position is a warning that humanity might be construction a new Tower of Babel. Technology, he writes, is neither an enemy of humanity nor something evil in itself; it is only ever "never neutral," because it quietly takes on the priorities of whoever designs it, pays for it and puts it to work.
That leads him to a deliberate piece of modesty. He declines to referee the technical questions, treats the human and social sciences as allies rather than rivals, and concedes that the Church holds no "definitive opinion" on matters that properly belong to researchers. Its authority, as he draws the line, covers ends and meaning and stops at the machinery.
What he keeps for himself is the older question the engineers tend to hurry past: what, in the end, are the machines for?
Key figure
945 TWh
Projected global data-centre electricity use by 2030, roughly double the level of the mid-2020s (IEA / UN University).
Many of the encyclical's claims are the sort a science desk can actually check, and those tend to hold up. Take its complaint about how much energy and water the technology swallows (par. 101). On the current evidence, that worry looks well founded.
UN University, drawing on International Energy Agency figures, expects global data-centre electricity use to roughly double, to something like 945 terawatt-hours a year by 2030. Most of it will go on the billions of ordinary queries that follow a model's release, the unglamorous part of the business.
Where the Pope and the Programmers Agree
On the narrow question of what AI actually is, the encyclical and the laboratories turn out to be closer than either side might care to admit.
Leo writes that a system can imitate a human being while possessing none of the things that actually make one. It has no moral conscience, no empathy, nothing of our relational or spiritual life. A machine can be made to sound a great deal like one of us without ever quite being one.
...we must avoid the misconception of equating this type of “intelligence” with that of human beings. These systems merely imitate certain functions of human intelligence.
Pope Leo XIV, Magnifica Humanitas
By 2026 the science had drifted to roughly the same place on the present tense, if not the future one. No current system shows anything the field would accept as evidence of sentience, though whether one ever could remains genuinely unsettled.
The two arrive at that agreement from almost comically opposite directions. The Church reasons downward from a soul it is sure no machine can hold; the labs reason upward from the awkward fact that no test they have built can locate consciousness in anything at all. They concur about the machine, if about little else.
The agreement reaches, of all places, the battlefield. The encyclical's hardest sentence is about weapons: "there is no algorithm that can make war morally acceptable" (par. 199).
Hand the decision to kill over to software, Leo argues, and you lower the threshold for using force while quietly reducing the people at the other end to data. He pushes this further than most clergy would dare, suggesting the old machinery of "just war" can no longer hold a conflict fought at machine speed (par. 192). It is a point arms-control and AI-safety researchers have been making, in their own idiom, for years.
The Power Problem Underneath It All
For all the theology, the worry the encyclical keeps circling back to is a thoroughly worldly one. The tools, the data and the decisions, Leo notices, are collecting in very few hands.
A more moral AI will not rescue us, he writes, "if that morality is determined by a few" (par. 107). The question he keeps returning to is one of ownership. Who, once the machine works, gets to hold the controls?
The anniversary is doing more than decorum here. The first Leo faced the factory and the working class it threw up; this one faces the digital economy, and frets that AI tends to set the pace of work and leave the human scrambling to keep up (par. 150).
Read in full, much of Magnifica Humanitas sounds like a labor charter for the algorithmic age: page after page asking for accountability across a system's life, for independent oversight, for an education worthy of the name.
The Real Argument Is About Limits
The real disagreement, when it eventually surfaces, has little to do with what AI can manage and a great deal to do with what we ought to want for ourselves.
A good stretch of the third chapter is given over to arguing, patiently, against that ambition. For Leo, those limits sit closer to the point of being human than the transhumanists will allow. They are the conditions under which relationship, love and openness to another person become possible at all.
What is
Transhumanism
A movement that treats human limits, including illness, aging and death, as engineering problems to be solved. Its aim is a "more than human" future built through technology rather than accepted as given.
On this one point his confidence is absolute. Human uniqueness, the encyclical says without flinching, is settled, secured by the conviction that each person is made in the image of God.
The consciousness researchers are nowhere near as certain. They will grant readily that no machine is a person today, then hold the door open, sometimes uneasily, to the possibility that one might be tomorrow.
Humanity flourishes not despite limitations, but often through them.
Pope Leo XIV, Magnifica Humanitas
That open door has already changed how the labs behave. Anthropic has set up a formal research program into the welfare of its own models, on the slender chance that there is something there that could be harmed.
Google DeepMind, for its part, has been hiring researchers to study machine consciousness head-on. Even those who take the question seriously rate the odds low for anything running today, while watching their own estimates climb sharply as they look into the coming decades.
The companies that built these systems are no longer willing to say, with the Pope's certainty, that the matter is closed.
What Both Sides Will Be Watching
Leo closes the document by asking, in effect, for the argument to carry on.
More On AI vs Humanity
Humanity's Last Exam: The Hardest AI Benchmark Yet
AI benchmarks are meant to measure what machines can and cannot do. But the machines keep outsmarting the tests. A new test tries to solve…
→When he frames what is at stake, he waves away the tired choice between cheerleading and panic. The real decision, as he puts it, lies between two kinds of progress: one that serves people, and one that slowly bends them to a logic of power (par. 129).
Religion and science, on his telling, are meant to work out together which of the two prevails.
He has also put a little institutional weight behind the sentiment. In May 2026 the Vatican set up an Interdicasterial Commission on Artificial Intelligence, which at least suggests it means to argue from inside the room rather than from the steps outside.
Which leaves the two camps roughly where the encyclical found them. They agree, more or less, on what today's machines are not. Their quarrel is over tomorrow, and over the very old question that the very new one keeps dragging back into the light: what, after all of it, a human being is actually for.
Sources
- Primary: Magnifica Humanitas, Encyclical Letter of Pope Leo XIV (The Holy See, 2026)
- Context:
- Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints (UN University, 2026, drawing on IEA data)
- Pope Leo's Magnifica humanitas: AI must serve humanity not concentrate power (Vatican News, 2026)
- The carbon and water footprints of data centers and what this could mean for AI (Patterns / Cell Press, 2025)
Fact Check: Claim-by-Claim Verification Mostly verified
All major factual claims about the encyclical are accurate; two errors were found and corrected — the wrong Anthropic co-founder was named as present at the Vatican launch, and a pullquote cite contained a spelling error in the encyclical title.
Editor note: The article originally named Dario Amodei. Corrected to Chris Olah — Anthropic's own post confirms Olah was the attendee, not Amodei.
Editor note: Pullquote cite typo "Magnificia" corrected to "Magnifica."
Commentary
- The claim "no current AI system shows evidence of sentience" is a broadly accepted scientific view but is framed here as editorial commentary — no single authoritative source is cited, which is appropriate for a scientific consensus statement.
- Direct quotes from the encyclical text (par. 107, 150, 192, 199) are paraphrases or close renderings; exact wording should be verified against the full published Vatican text before final publication.
- The article's framing of Rerum Novarum as a "letter on factory labor" is a simplification but not inaccurate; the encyclical addressed industrial labor conditions broadly.
Sources used for verification
Academic / Primary Sources:
- Magnifica Humanitas — Vatican (Holy See)
- Environmental Cost of AI's Energy Use — UN University / IEA
Other reliable sources:
- Magnifica humanitas: AI must serve humanity not concentrate power — Vatican News
- Chris Olah at the Vatican encyclical presentation — Anthropic
- Pope Leo, Anthropic co-founder call for Church-tech ethics partnership — National Catholic Reporter
- Pope Leo XIV and Villanova University — Villanova University
- Exploring Model Welfare — Anthropic
- Google DeepMind hires philosopher for machine consciousness research — LinkedIn
Fact-checked by Perplexity Sonar Pro on 2026-06-09
