The a[i]theist who needed the Pope

I do not believe in God. I have never felt a God-shaped hole in my life, which is convenient, because I would almost certainly have tried to fill it with books, coffee, road miles and a badly timed argument about quantum physics. I’d rather listen to Christopher Hitchens than to a priest preaching a creation in seven days while sculpting whole women out of a man’s rib. The thought alone.

I am not spiritually searching. I am not hovering at the edge of conversion. Nobody needs to send candles. I have never needed one. I am the kind of atheist who finds the religion questions mildly boring, the way some people find astrology mildly boring, with the added irritation that astrology has rarely started wars, burned witches or decapitated heathens. So you can imagine my professional discomfort on Monday 25 May 2026, sitting in my o-so-comfy office in Aalter with a coffee going cold, reading an encyclical from Pope Leo XIV and realizing, slowly (I’m not that fast anymore), with the dignity of a man caught laughing at the funeral of an old friend, that he had written one of the sharpest documents on artificial intelligence I have read all year.

Magnifica Humanitas. “On safeguarding the human person in the time of artificial intelligence.”

A title that should belong to a McKinsey deck. Instead, it sits pontifically (see what I did there?) on Vatican letterhead, in eight languages, next to Rerum Novarum, the 1891 encyclical where Leo XIII looked at industrial capitalism and concluded, without the help of a searchbot, that workers were human beings rather than expendable industrial furnace accessories. Leo XIV is now doing the same move for AI. The machine is the weather system; the subject of the encyclical is what happens to the humans living inside it.

I keep wanting to dismiss this. I cannot. That is a personal embarrassment. Forgive me Hitchens, the force is strong in this one.

The scene that sets the scene

In the Synod Hall last Monday, Pope Leo XIV thanked Christopher Olah, co-founder of Anthropic, for accepting the invitation to come and talk about AI. A man in white, busy being infallible, surrounded by ecclesiastics, addressing a frontier AI researcher who builds Claude for a living. I thought I had seen it all. Olah then stood up and said every frontier AI lab, including his own, “operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing.” Who knew? Commercial survival, race pressure, geopolitics, pride, head-spinning ambition. He said the world needs “informed critics who will tell the labs when we are failing,” and “moral voices that the incentives cannot bend.” He said this in the Vatican, out loud, in front of, indeed, a Pope.

Well, hell is freezing over. A co-founder of a hundreds-of-billions AI company walked into one of the oldest institutions on Earth and said, in effect: please keep telling us truths our incentive structure will try to metabolize, soften, postpone or ignore. Please supply the moral voices this industry cannot.

Holy purple cow. This is a man at the controls of one of the most consequential technologies of the century saying, on the record, that the controls and the guardrails are not enough.

I am an atheist. I do not need a Pope. I do not need a god. I do not need a confessor or a saint or a calendar of holy days. Yet here I am, on a Wednesday afternoon, writing the sentence I never thought I would write: somebody has to tell the labs the truth, and the labs have just admitted they cannot do it themselves, and the only institution available and ballsy enough, with two thousand years of practice at saying uncomfortable moral sentences out loud, happens to wear a cassock.

I can hear Christopher Hitchens whispering in my ear: “mind, the Church has been wrong about a lot of things.” And it has been, and it still is. Fine. Welcome to 2026. The materialists became the mystics. The Church became the empiricist. Nothing is where you left it.

Silicon Valley does theology all day

It just hates admitting it. Salvation is AGI. Damnation is extinction risk. Confession is a model card. Prayer is a prompt. Saints are founders with podcast microphones (you know the ones). Monasteries are hacker houses with worse hygiene. The eschatology comes with a Series C and a ketaminated Stanford dropout.

I do not say this as a sneer. I say this as a man who has sat in too many conferences where a thirty-year-old with a great corporate haircut, a Rolex and $80 million in funding explained to me, with the facial semi-holy expression of a minor prophet, that “frontier intelligence” would soon “transcend” the human condition. The phrase was always delivered the same way: voice slightly lowered, eyes slightly widened, as if the speaker had been allowed to read a passage of scripture early. Nobody ever flinched. Nobody ever said: this is theology with an obese venture capital chassis and bad breath.

Then Rome walks in. Excuse me, the Vatican waltzes in, Pope and all, and perversely, gives the materialists a lecture. The Pope talks about data centers, water, energy, miners, content moderators, labor dignity, platform concentration, kill chains, children crushing rare-earth ore. The lab people talk about mystery, inner states, model welfare, joy, grief, geopolitics, and whether the machine might one day be a moral patient. The Church brings bodies. The labs bring metaphysics. I had to put the laptop down and look out the window for a minute. Am I still in the right movie?

The Babel sentence

So I read the thing. Magnifica Humanitas is an intriguing, powerful read with a parable choice between Babel and Jerusalem. One city built around a single tower, a single language, a single project, a single fantasy of control, a single “X” (pun intended, in your face Musk), designed to “make a name” for itself. The other rebuilt piece by piece by Nehemiah, families and communities sharing the work, repairing what had collapsed. I rolled my eyes when I read this. Bible cosplay. Move on.

I was wrong. Babel is maybe a good mental image. AI’s real danger goes well beyond ChatGPT writing bad poems (though a plethora of tough crimes have been committed). The danger is the dream of a single digital grammar that can translate a human being into data, performance, probability and price. One model, one stack, one ranking. One inference pipeline through which everything passes and from which nothing escapes. Rome calls that a temptation. The industry calls it transformation. I know which one is more honest.

The Pope writes that “the primary choice is not between a ‘yes’ or ‘no’ to technology” but between building Babel or rebuilding Jerusalem. Between a power that claims to dominate the heavens, and a people who repair the walls of fraternal coexistence in the presence of what they hold sacred (even if that word gives me the flying hillbillies).

Luckily, you do not need to believe in that sacred to feel the force of the sentence. You only need to have sat through a single quarterly review in which “scaling” and “rightsizing through technology” was treated as an end in itself.

The AI-detector scandal, taken seriously

Now, the comedy (it’s about time, said Dante). The Verge reported that AI detectors flagged portions of Magnifica Humanitas as possibly AI-assisted. Linch Zhang at LessWrong ran the text through Pangram, the best commercially available detector, and got between 40% and 100% on certain paragraphs. The Verge ran roughly 2,000 words through Pangram and got 46% AI. The first chapter clocked 62%. Pangram claims a false-positive rate of about 1 in 10,000.

Cue a thousand smirking posts from keyboard warriors whose own moral philosophy was last updated by a Joe Rogan clip. But honestly, who cares? Even I cram my writings through the throat of Grammarly, and Pangram told me one day a 2006 blog post was 63% AI. I know, I always was a little bit ahead. Bon, let us take the irony seriously.

If AI touched the drafting, the scandal is not merely hypocrisy. The scandal would be opacity. Where did assistance end and authority begin? Who reviewed? Who signed? Who owns the consequences of the words? Do they stay infallible if AI touched them? (I know.) That is precisely the line the encyclical draws everywhere else: assistance is not authority, drafting is not discernment. A tool can help produce language; a human institution still has to answer for the language it releases. That is the corporate AI governance problem in a frock.

When a consultant uses Claude to draft strategy, who owns the advice? When HR uses AI to score candidates, who owns the rejected life? When procurement uses AI to rank suppliers, who owns the hidden bias? When a military system accelerates target selection, who owns the dead? The answer cannot be “the model.” That sentence is cowardice with better latency.

So yes, an atheist can read the Pope on AI and laugh at the detector results.

“Disarm” is the word that broke me

The Pope uses the word “disarm.” He does not mean smashing GPUs with a chalice (though some vendor demos do test one’s ecumenical restraint). He defines disarmament as freeing technology “from the mentality of armed competition” (wow), which today is not just military but economic and cognitive: the race for ever larger algorithms and ever larger datasets, driven by the desire to secure geopolitical or commercial dominance. Then the sentence that should be bolted to the wall of every AI lab, every regulator, every boardroom, every defense ministry:

“To disarm means discrediting the assumption that technical power automatically confers the right to govern.”

Yep, right on, Leo. Silicon Valley behaves as if ownership of compute implies jurisdiction over the future. Sam, Sundar, Satya, Mark, Elon, Dario. Six men. Six decision graphs. A combined capex this year that overshoots the GDP of most countries that will have to live with the consequences of their decisions. They have the compute. They believe they therefore have the say. We’d rather they did not. And Rome just said no.

Mind: not elegantly, not technically, not with a cowardly benchmark. Nope. With a much older category: authority. The Pope: “Authority without moral formation becomes domination.” (And yes, this should definitely not come from a Pope who reigns over the value scheme of way too many people like a moral despot.) I am an atheist. I do not need that framing to be true. I would have preferred to find it in a working paper from Brookings, or a NIST publication, or an EU Commission communication. Those exist. They are necessary. But, let’s be honest, they are also nowhere near as quotable. The Pope wrote a sentence that a chief risk officer can actually deploy in his next board meeting. That is (witch)craft.

Bodies under the cloud

The encyclical is at its sharpest when it drags AI back into the body. Nothing in AI is immaterial or magical, Leo writes. Every instant response is the end of a long chain of mediation: energy, water, (stolen) IP, infrastructure, networks, and “above all, people.”

He even names and lists them: data labelers. Model trainers. Content moderators wading through disturbing material on minimal wages. Often young. Often women. He names the rare-earth supply chain. He names children and adolescents working in dangerous conditions, crushing the ore from which the rare earths are extracted. “The bodies of these people are scarred, injured and worn down so that computational flow may continue uninterruptedly.”

That sentence is missing from every launch video with a glowing orb and an executive saying “democratize.” The cloud is not a cloud. It is a mine, a grid, a cooling tower, a content-moderation queue, a water bill in Maricopa County, a trauma pipeline in Kenya, and a financial instrument with a nicer investor-friendly interface.

Again, the spiritual leader is the one pointing at matter. The supposed materialists are selling destiny. Lovely.

Data colonialism, named directly

Leo names it. Out loud. Without softening. “Dans la gueule,” as the French say. He writes that colonialism today “no longer dominates only bodies” but appropriates data, “transforming personal lives into exploitable information.” Entire regions, especially those “marked by structural fragility and limited geopolitical relevance,” are subjected to a new extraction regime: health data, epidemiological profiles, genetic maps, demographic information. He calls these the “new rare earths of power.” He notes that this data is “often collected under the pretext of aid, research or innovation.” Those last six words are the operating model of half the global health-tech industry.

The verdict: if the world does not return to individuals not only the data that describes them, but the right to decide how it is used, by whom and for whose benefit, “the digital age will not be post-colonial, but colonial in another form.”

The Pope just named the business model with its jacket off and its pants down: stealing. In a boardroom, that sentence would have triggered a slide reshuffle. In an encyclical, it triggered three press releases and a thousand (very angry) op-eds.

Anthropology, the word I keep avoiding

I hate using the word “anthropology” because it sounds like someone just lit a candle in a policy seminar for retiring archaeologists. But here it is, the uncomfortable noun the previous version of this piece kept fumbling. Every AI system carries an anthropology.

A hiring model carries a theory of potential. A pricing model carries a theory of value. A classroom tool carries a theory of learning. A companion bot carries a theory of loneliness. A defense system carries a theory of whose face can be abstracted into a target.

Philosophy ceases to be decoration and becomes a weapon. This is architecture. This is vendor selection. This is the contract clause nobody reads because the meeting is running over and someone has a train to catch. The Pope writes: “ethical discernment cannot be limited to asking whether we are using a system for good or bad purposes; it must also examine how that system is designed and what vision of the human person and society is embedded in the data and models that guide it.”

If I had read that sentence in a Brookings paper I would have nodded and moved on. Reading it in an encyclical, an atheist’s nodding is louder, because it is also surprised. Hitchens would have uncorked some more bourbon.

War, and defense procurement

“No algorithm can make war morally acceptable.” Are you listening, Pete Hegseth? The encyclical also says lethal or otherwise irreversible decisions cannot be entrusted to artificial systems. It demands identifiable, verifiable human responsibility for those who design, train, authorize and use such systems. It says cyberspace is now a battleground and needs negotiated international rules to protect civilians from “invisible yet real forms of violence.”

Will defense (or war) procurement be haunted by this? Of course not. Procurement has a remarkable immune system against haunting. Still, the sentence will outlive a hundred white papers. Quotable matters; the Church is good with words. The Geneva Conventions of 1949 are quotable. The ICRC’s 2021 position on autonomous weapons is quotable in committee. “No algorithm can make war morally acceptable” is quotable in a parliament, on a placard, in a chief risk officer’s email, on the front of a newspaper.

That is what an encyclical is for. It launders a moral claim into a sentence that can travel. Respect. I recognize art when I see it.

The Tara paragraph

For those still reading: respect. This is where I slow down a notch or two, because Tara is ten, and I have zero interest in turning my own kid into a convenient prop for internet solemnity. The encyclical’s education section says something simple and hard: the speed and ease of AI answers can extinguish the desire to ask questions, and schools should offer what the digital sphere cannot provide, namely “a shared time for learning and developing trustworthy relationships.”

Silence is now countercultural. Patience is countercultural. A child asking badly, circling back, getting frustrated, trying again, misunderstanding beautifully and growing a thought by friction is countercultural. The machine offers the answer before the question has done its work. Adults love this because adults are tired (I am tired), and because we have trained ourselves to confuse completion with understanding.

Tara deserves better than autocomplete childhood. So do adults, frankly, though some of us are probably beyond warranty.

The Church’s own ledger

The thing that nearly knocked me out of my chair is that the encyclical does not arrive innocent.

It talks about slavery. Bluntly. It notes that the Church’s formal, absolute, universal condemnation of slavery came only in the 19th century, “notably under Pope Leo XIII.” It notes that in antiquity and the Middle Ages “many individuals and even ecclesiastical institutions had slaves.” It uses the words “wound in Christian memory” and “in the name of the Church, I sincerely ask for pardon.” It also talks about ecclesial abuse, listening to victims of “spiritual, economic, institutional, sexual and power-based abuse,” and the need for transparency, reparation, and structural reform inside the Church itself. That language would have gotten a content grunt from the Hitch. Rome is not speaking from innocence here. It is speaking from a long institutional record of learning moral facts late, often after bodies have paid the invoice.

That, oddly, is the most useful warning in the document: do not make us apologize in 2126 for the digital colonialism we can already see in 2026.

So why does an atheist need this?

I do not. For fuck’s sake. And I do. I do not need a Pope to tell me that data centers burn fresh water, that miners are children, that hiring models encode bias, that defense systems should not pull triggers, that schools should be slower than TikTok. I have read Kate Crawford, I have read Shoshana Zuboff, I have read the EU AI Act line by line because that is my job. The substance is not new.

What is new is the placement. When Brookings says it, three CIOs read it. When the OECD says it, ministers read it. When the EU Commission says it, lawyers read it. When the Pope says it, a billion people hear it, half the boards in Europe quote it on Friday, and ten thousand journalists who would never read a working paper write something almost true about it on Tuesday. That is power. Pretending otherwise is amateur.

I am an atheist. I will remain an atheist tomorrow morning, and the morning after, and on the morning of my own funeral if anyone is brave enough to schedule it. I do not need a god. I do not need a saint. I do not need an afterlife. I do not need an encyclical.

I need a sentence that can travel. Magnifica Humanitas is full of them.

The boardroom version

If you run a company in 2026 and you are using AI in recruitment, customer scoring, credit, pricing, risk, healthcare access, education, insurance, policing-adjacent analytics, public services, or anything involving children, you do not need to convert. You need to read paragraph 109. “Social justice must shape the very design” of these systems “from the outset,” not be retrofitted after deployment. That is not theology. That is product policy. It is also almost verbatim what we put in the IEEE paper Ethically Aligned Design: A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems almost a decade ago.

Find the systems already making or shaping consequential judgments. Not the shiny AI strategy demo. The real ones. The HR scoring tools, procurement recommenders, customer-service summarizers, finance anomaly detectors, sales agents, pricing optimizers, knowledge-base goblins, browser extensions, shadow workflows and vendor features that arrived quietly during a product update while everyone was pretending governance lived in a SharePoint folder.

Follow the recommendation until it becomes a decision. Follow the data until it becomes a person, with a name and a postcode. Follow the interface until it becomes authority. Follow the contract until responsibility evaporates into a vendor T&C nobody read since onboarding.

Follow the silence until you find the human who cannot appeal.

That is where Magnifica Humanitas belongs. Not in a Vatican-themed AI ethics panel with polite applause and holy mineral water. On the table in risk committees, in procurement, in agencies, in schools, in ministries, boardrooms and war departments. In the CIO’s inbox, preferably before the next enthusiastic business unit launches a semi-autonomous agent called Sophie and gives it access to customer data because someone thought the onboarding video was cute.

The embarrassment

The AI industry wanted to build tools so powerful they could reshape civilization. Congratulations. Civilization has opinions. The Vatican voiced those.

The Pope did not solve AI. He did something more irritating. He made evasion harder. He looked at the labs, the governments, the boards, the vendors, the believers, the doomers, the accelerationists, the regulators, the parents, the teachers and the rest of us conveniently busy people, and he asked the question the whole circus keeps trying to outrun: Who is this for, and who pays?

I am an atheist. I still do not believe in God. I still do not need a god. I still do not need a Pope. I needed that question. I needed someone with a microphone loud enough to land it where it had to land.

Rome noticed the bodies under the cloud. That is the embarrassment.