Catholic AI Development, Human Flourishing & AI Builders

At the GRAiCE MED Global Summit, convened by Georgetown University’s AI CoLab in Rome on Wednesday 23 September 2026, Matthew Harvey Sanders opened on a vow seven hundred years old. Brothers joining the Hospital of Saint John of Jerusalem promised to be “the serf and slave of our lords the sick”. His argument in “Catholic AI Development, Human Flourishing & AI Builders”: skill can be handed to a machine, and a vow can’t. Read the full talk below.

I. Our Lords the Sick

More than seven hundred years ago, a man joining the Hospital of Saint John of Jerusalem came before the Chapter on a Sunday, put his hands on the missal, and made the three promises every religious makes. Obedience, chastity, poverty. Then he was told there was a fourth. The words survive in the order's own book of customs: "we make another promise, which no other people make, for you promise to be the serf and slave of our lords the sick."

Our lords the sick. That was the hospital's name for its patients. Its rule, older still, required that a sick man be carried to bed and fed "as if he were a Lord," every day, before the brothers themselves sat down to eat. Many of those brothers were knights, the sons of knights, who'd spent their lives being served. The order that kept that hospital still exists. It's the Sovereign Military Order of Malta, and it governs itself from two addresses in this city, a palazzo on Via Condotti and a villa on the Aventine.

That's the tradition the hospital itself comes from, and I'll come back to it, because it's about to become the most practical idea in medicine. The vow never asked how much a brother knew. It asked whose he was. Skill can be handed to a machine, and most of it is about to be. A vow can't. That's the difference this talk is about.

I'm not a physician. For the past several years my work has been making two thousand years of Catholic writing readable to machines, and then deciding what a machine should do with it once it can. The organizers asked me to tell you what we built and what we decided. One of those decisions is now yours to make too, and it only makes sense against the scale of what's coming. So let me start there.

Demis Hassabis is a Nobel laureate and a co-founder of Google DeepMind. In 2024 Pope Francis appointed him to the Pontifical Academy of Sciences. He describes himself as a cautious optimist. He told The Guardian in August of last year: "10 times bigger than the Industrial Revolution," he said, "and maybe 10 times faster."

Pope Leo XIV has given the period a name. In Magnifica Humanitas he calls it "the age of AI and robotics." Notice the second word, because nearly everyone leaves it out. Every previous wave of automation took muscle and left the mind standing; people moved into knowledge work. This time AI systems are taking cognitive work while robots take physical work, inside the same decade, and there's no third category to move into. Unless there is a third kind of work, which was never muscle and never knowledge: being present to another human being. Hold that thought.

The cognitive half is already in your imaging, your triage and your documentation. The physical half has arrived too. At Taichung Veterans General Hospital in Taiwan, a nursing robot called Nurabot already carries medication, specimens and wound care kits to the bedside. The World Health Organization projects a shortfall of about eleven million health workers by 2030. These machines aren't arriving in your hospitals as a threat. They're arriving as relief.

And a robot that carries the medication is the modest beginning. The ambition runs a great deal further. Elon Musk, whose company builds humanoid robots, said this at an investment conference last November, and I'll quote him at length, because he's usually quoted in fragments.

"Imagine a world where everyone has access to the best surgeons, literally everyone. And Optimus will have the level of precision that is frankly superhuman, and will be able to do medical procedures, very sophisticated medical procedures, any medical procedure, perhaps things that really humans can't even do because they're too difficult, and that will be available to anyone."

Then he named the constraint he believes this removes. "People often talk about eliminating poverty and providing great medical care, but they never actually have a solution." Money doesn't solve it, he said, because "there's a very limited number of great doctors and surgeons. They don't grow on trees, but now they'll get built in factories."

He has said three years; I think it'll take longer. But set the date aside and look at what he's pointing at, because he's right about it. Great physicians have never scaled. It takes more than a decade to make one and they can only be in one room. Every health system in history has been designed around that constraint, and it's why the finest care on earth has always been rationed by geography and luck. A child born in one country gets a specialist; a child born four hundred miles away gets a clinic with no anaesthetist. Nothing has ever changed that. This might.

So, the optimistic case, plainly. Diagnosis, surgery and treatment are about to become available at a scale and a quality that human history has never seen. Enormous amounts of suffering are going to end because of it. A Church that has prayed for the sick every day for two thousand years should be the first to welcome that.

II. What Remains When the Machine Is Right

Now look carefully at what Musk is describing. He's more precise than his critics allow.

He's talking about the technical work of medicine, the parts that are in principle mechanical. He's not saying that a human being at a bedside stops mattering. He isn't thinking about the bedside at all.

Asked in January whether young people should still go to medical school, he said no, and called it "pointless." That's where I think he's wrong.

Because the diagnosis and the procedure were never the whole of the doctor. They were the part we knew how to measure.

There's the consultation where the diagnosis is right and the patient hears nothing after the first sentence. There's knowing which of two families needs the news slowly, and which needs you to stop talking. There's the judgement to stop treating. There's noticing that the man in front of you has answered every question correctly and isn't going to take the medication. There's sitting down instead of standing up. And there's owning the decision, the one you take home and lie awake with at two in the morning. No machine will ever lie awake.

None of that is mere sentiment. It's clinical work that no protocol fully captures, which is why it was under-measured and never got the standing the technical half got. You've known for decades that presence is a clinical variable. It was simply never the one the system was built to reward.

And it's about to be the half that's left. When the diagnosis comes from a machine and the procedure is done by one, what a physician is for becomes a question about presence.

I don't mean that as a consolation. I mean it as a job description. The doctor who has the machine's diagnosis in her hand, and the forty minutes she used to spend reaching it, isn't a diminished doctor. She is the first doctor in generations who can afford to sit down. Whether she does, and whether anyone trains her to, is the question your profession is about to answer. Because those forty minutes won't stay empty. Either your profession claims them for the patient, or someone else will claim them for the ledger and call it efficiency.

What I can tell you is that this exact question, what the human being is for once the machine can both know what a physician knows and do what a physician does, is one my company has been building around for years. And what we learned is usable in this room.

III. The Bridge, Not the Destination

When people hear what my company does, they assume we built a machine to take the place of the people who used to answer questions about God.

We didn't, and the difference is the whole talk.

The Catholic Church carries two thousand years of writing. The Fathers of the early centuries. The Doctors, Augustine and Aquinas and Teresa and the rest. The councils. The popes. That body of work is, among other things, a two-thousand-year inquiry into what a human being is and what it means to care for one.

The large models have read most of it. That isn't the problem. The problem is how a model holds what it has read. Training compresses a library into a memory, and that memory is uneven. A model can sometimes reproduce a passage word for word, usually one it has seen many times. It can't do so reliably, and it can't tell you when it hasn't. Where recall fails, it produces something plausible instead, and that is a hallucination. So the question we set ourselves was never how to make a machine sound Catholic. It was how to put the actual text in front of the model at the moment of the question, so that the answer is read from the source rather than remembered, and can be cited.

That's the mission, stated plainly: Catholic AI, at scale. In practice it means digitising everything the Church has written and putting it within reach of anyone in the world, on any device, in their own language, at the moment they need it most. Not only the scholar who can get into a Roman library. Anyone.

It took three pieces, in order, and the first wasn't a chatbot. It was the Alexandria Digitization Hub, which we run in this city with the Pontifical Gregorian University, to make that inheritance machine-readable. Robotic scanners processing thousands of pages an hour. Among the first things through them was the Magnum Bullarium Romanum, thirteen centuries of papal documents, from Leo the Great to Benedict XIV.

The second piece was a platform called Vulgate AI, once we discovered that the harder problem was never scanning. Much of what institutions have scanned is still invisible. A scan is a photograph of a page. A machine can no more search it than it can search a photograph of a filing cabinet. It has to be turned into text a system can actually retrieve, and few institutions have done that at scale. I'd expect a century of case notes in your own institutions sits unretrievable too.

Only then did we build the third piece, Magisterium AI. Someone asks a question about suffering, or a moral decision, or whether God is there. The system answers from the actual texts, with citations, so the person is reading Augustine's words and not a paraphrase. A widow asks how anyone survives grief, and what comes back isn't a machine's sympathy but Augustine on the death of his own friend, sixteen centuries ago, with the reference, and beneath it the encouragement to find a parish and a community of people who've stood where she's standing. Over a million people use it now, across a hundred and ninety countries and more than seventy-five languages.

What we're doing is using artificial intelligence to connect people to the Doctors and Fathers and Popes and heroes of the faith. To let those men and women fight for people's souls again. The model is the bridge. It was never meant to be the destination.

Most people won't leave the assistant they already use to ask a question about faith. They've built their working lives inside ChatGPT or Claude or Gemini, and that's where they'll ask. So we built connectors that put Magisterium AI inside those assistants: ask about the faith with our connector in place, and the answer comes back from the Church's own sources, cited, without changing apps. If the model is a bridge, it shouldn't matter which bank you start from.

Not everyone wants to talk to a machine about their soul, and they're entitled not to, so two of our tools involve no conversation at all. Search, where you type what you're looking for and read the original document yourself. Explore, the whole body of Catholic writing laid out as a three-dimensional map, so you can see what sits next to what. Neither talks to you. Both hand you the Church's own words and get out of the way.

There's a fourth project, called Ephrem. It's a specialised language model, the first Catholic AI trained from scratch rather than adapted from somebody else's, and it's small enough to run on hardware a family owns, so the most private questions never have to leave the house. So that no one leaves with the wrong impression: it's unreleased, it's research, and we're aiming at 2027.

And across all of it we made one decision that I think transfers to your field.

We built the system to point beyond itself.

When someone gets their answer, the goal is that they close the laptop and go to a parish. That they find a religious community. That they sit across a table from another human being and have the conversation in person. We'd sooner lose the session than keep it. It's the opposite of how almost every consumer AI product on earth is designed. The hard engineering isn't the answer. It's recognising the moment when a person has stopped needing information and started needing someone, and knowing where in their own city to point them.

Now the objection, from inside the Church, which I take seriously. Last November a Catholic critic published an essay called "Beware of Catholic AI," and named us in it. Her strongest points, in her words: "the Holy Spirit does not guide the outputs of LLMs," and "prudence cannot be produced as output; it must be lived."

She's right. And her objection is aimed at exactly the misunderstanding I started with. She's afraid we built the machine to be the thing people relate to. But the machine carries people to Augustine and then tells them to go and find a living Christian. It was designed on her premise.

Here's the transferable principle, and it's a procurement question. When someone sells you a patient-facing system, ask what it does at the moment the right answer is a person. Ask whether it's built to hold the patient or to hand them on. Ask what it measures as success. If the honest answer is time on the platform, you're buying a product whose purpose isn't your patient. The technology to build the other kind exists. What's missing is usually the decision.

The Pope has said it more directly than my industry ever has. This is Magnifica Humanitas, section 170: "When business models thrive on human weakness, the person is treated as a means rather than as an end; those who design or finance such systems bear a moral responsibility that cannot be ignored." He is describing my industry, and he is right. So the question for anyone who builds these systems comes down to the one he asks: in your design, is the person a means or an end? Everything I've described this afternoon is our attempt to answer: an end.

IV. The Builders and the Vow

That decision is easier to make when you're not making it alone, which is why I was asked to speak about the Builders AI Forum.

Last November about two hundred of us met here in Rome, at the Gregorian, to ask what it means to do this work as Catholics. Engineers from Microsoft and LinkedIn, venture capitalists, media producers, and people from health systems, CHRISTUS Health, the Bambino Gesù, sat in six working groups with bishops and Vatican communications officials, each group handed one specific challenge. Not to write another ethics framework. To compare what we'd actually shipped, and what we'd refused to.

Pope Leo XIV sent the forum a message. One sentence has stayed with me: "every design choice expresses a vision of humanity."

You can read a specification and work out what its authors believe a person is. He went on to ask builders to make moral discernment a fundamental part of their work, and said the work "cannot be confined to research labs or investment portfolios." Then he gave three examples of the work he had in mind. One of them was "tools for compassionate healthcare."

I was in that room, and this summit is the same conversation with more clinicians in it. The forum's real product isn't a document. It's a group of people inside the industry who hold each other to that decision. That's how the ethos of a field changes: not from outside, but by enough builders deciding together what they won't optimise for.

The Hospitallers knew this long before us. Their rule fit on a few pages. What kept it alive for centuries wasn't the ink. It was a promise made aloud, in front of men who'd made the same one.

V. Body and Soul Together

Which brings me back to Jerusalem.

The Hospitallers' statutes of 1182 required "four wise doctors," qualified, in the text's own words, to examine urine, to diagnose different diseases, and to administer appropriate medicines. They wrote rules about the beds and the sheets and the bread. They cared about outcomes, and by the measure of their age the care was extraordinary. But that isn't the most important thing that set the place apart. What set it apart most was the view of the patient the brothers carried into the ward, and the formation that put it there: a religious life, the hours prayed and the rule kept, inside a tradition that taught them the worth of a human life. That formation is why they could call a dying beggar their lord and mean it. Body and soul were treated together because the people doing the treating had been formed in both.

That's what Catholic medicine is supposed to be, and I'll say gently that the hardest thing about Catholic healthcare today is telling it apart from everyone else's. Same reimbursement codes, same throughput targets, same fifteen-minute slot. A patient can walk into a Catholic hospital and walk out having met a very good system and no one in particular.

Two things follow from Jerusalem, and the machines make both possible at a scale they never were.

The first is about what the system is for. In Jerusalem the hospital existed for the patient; the brothers ate after the sick had eaten. Now look at where the money goes. The OECD's latest figures, published this March, are for 2023. Across its member countries, twenty-eight percent of health spending went to inpatient care. Three percent went to prevention. A system paid for treating illness has a quiet interest in illness to treat, and almost none in the patient who never needed it. Catholic medicine should be the part of the system that cares about that patient. Not because it's cheaper, though it often is, but because keeping a person whole, at home, and in their own community is what caring for the whole person means. And it's now possible at scale. A person's health can be watched over continuously, with their consent, on a device they already own, and the pattern that precedes a crisis can be caught days before it comes. When they do need you, the record arrives before they do, read by your model and by you, so the diagnosis is well underway before they walk in, and they're home sooner. With one rule: the data belongs to the person, and it moves only toward the people caring for them.

The second is about who the clinician is. The brothers were formed by the life they lived, every day, alongside the work. You had years of formation, and for most of you most of it was technical. The other half, the worth of a life, how to stay in the room with fear, what to do with your own grief at the end of a thirty-year career, was left to whatever you happened to bring with you. That's the half the machine is about to leave you holding. And the tradition that formed the Hospitallers is exactly what we've spent these years making readable. The Church's deepest minds on suffering, on fear, on staying present to the dying are in that corpus, and a physician who wants to become the kind of person a patient can lean on can put a question to Aquinas or Teresa or Newman between rounds. That's what we mean when we say these tools are for formation. The patient's, yes. But also yours, because it's hard to be a witness to something you haven't first been formed in yourself. The brothers had the hours of the day for that. You've got the corridor between two rooms. We built for the corridor.

The Hospitallers wrote their view of the patient into a vow, and put a whole way of life behind it. We should be willing to write ours into the code, and to put the formation back within reach of the people who have to live it.

VI. Only a Person Can Carry the Witness

I've been circling one idea since I began, so let me say it straight.

If a text alone could have saved us, God could have sent a text alone. A perfect one, in every language, for everyone at once. It's not what happened. He sent a life. A particular man in a particular town, who touched people and was touched, who sat down and ate with them, who died, and who rose. He left a witness, and told the witnesses to go and be witnesses.

That's the reason we built the machine to point beyond itself. Knowledge of the faith matters, and it's not enough. People come to believe that a thing can be lived when they see someone living it, and there's no substitute for that, not in a book and not in a model. A saint is simply a person whose life has become the evidence. Informed people, yes. But not only informed. Saints, whose lives other people can look at and think: so it can be done. The machine can carry the knowledge. Only a person can carry the witness.

And that's also what medicine is, at the moment it matters most.

The machines may one day end most disease. They won't cure death. Nearly every clinician in this room has stood at a bedside where there was nothing left to try, and the rest of us will be brought to one eventually by someone we love. The last option is gone and the family is in the corridor. No model improves that moment. What the patient needs is a witness: a human being whose presence says that they're still a person, that they're not alone, and that care is possible even when cure isn't. No machine can give that. Not because it isn't clever enough, but because it isn't anyone.

That's your embodied role. It isn't what remains after the machine has taken the real work. It was the real work, and the machine is about to make that undeniable.

Let me leave you with a doctor who understood it.

Giuseppe Moscati worked in Naples, down the coast from here, and died in 1927. When John Paul II canonised him he described him precisely: chief physician of a hospital, a distinguished researcher, a university lecturer in physiology and physiological chemistry. His kindness wasn't a private virtue that sat beside his medicine. He was a serious scientist, and he took that same seriousness into the homes of the poorest patients in the city. To a young doctor who'd been his student he wrote: "Remember that you must concern yourselves not only with the body, but with souls, through counsel, and by reaching down to the spirit, rather than with cold prescriptions to be sent to the pharmacist."

Set that sentence beside what's coming. The cold prescription will soon be produced perfectly, instantly, for nothing, by a machine. Everything Moscati set against the real work is exactly what we're about to automate. Which leaves the real work standing there, and a profession that has to decide whether to claim it.

His words again, as the Pope quoted them that day.

"Blessed are we physicians, so often unable to drive away an illness; blessed are we if we remember that beyond the bodies we have before us immortal souls, for whom the Gospel precept to love them as ourselves is urgent."

He said that when a doctor was powerless far more often than any of you are. He is about to be right again, for the opposite reason. Not because we can do so little, but because the machines will do so much, and what remains is exactly what he asked us to remember.

He put it in one line, which the Pope who canonised him called almost a testament, and I'll leave it with you, because he was a research scientist, and so are many of you.

"Non la scienza, ma la carità ha trasformato il mondo."

Not science, but charity, has transformed the world.

Thank you.