Artificial intelligence may be one of the greatest tools for evangelisation since the founding of the Church — or the moment the Church lost the narrative entirely. Matthew Harvey Sanders made that case to the Meeting of Press Officers and Spokespersons of the Council of Bishops’ Conferences of Europe (CCEE) at the Italian Bishops’ Conference (CEI) in Rome, 6 May 2026, arguing that the Catholic voice on AI is largely absent from public conversation, and that the Church’s communicators are the ones who must act.

---

SECTION I: THE DIGITAL RUBICON

Your Eminences, Excellencies, dear colleagues — and especially the men and women in this room whose work I have come to address: the press officers and spokespersons of the Bishops’ Conferences of Europe.

I want to start with what you do — not in the abstract, but in the concrete reality of your working week.

You are the people who translate the Church to the public. Every interview a bishop gives, every statement a conference issues, every pastoral letter that lands in a journalist’s inbox — somewhere along that chain there is one of you, shaping the words, anticipating the question, fielding the call at ten o’clock at night when a story is breaking. You are the institutional voice of the Catholic Church in Europe.

And the public you address is now, every single day, being formed by artificial intelligence. Not exclusively, not yet. But increasingly, and for the generation now being formed, primarily. AI is the newest and fastest-growing layer of information formation in the lives of the people you are trying to reach.

I will not rehearse the question of fear. I want to begin with confident action, because that is what your work requires, and what this moment demands of the Church.

We have crossed a threshold. Not an incremental one — a civilizational one. For roughly thirty years we lived in what we called the Age of Information. Machines retrieved, indexed, sorted. They found and organised what human beings had already written — powerful tools for retrieval, but not for reasoning. That age is over. We now live in the Age of Automated Reasoning. Machines no longer fetch — they generate, reason, and advise. They form judgments and shape consciences.

The Stanford AI Index released earlier this year sets out the scale precisely. Generative AI reached fifty-three percent of the global population within three years of its public release — faster than the personal computer, faster than the internet itself. Eighty-eight percent of organizations have adopted it. Four out of five university students now use it routinely. In the most recent Bentley-Gallup survey, thirty-one percent of Americans say artificial intelligence does more harm than good to society. Only thirteen percent say it does more good than harm. The people who must live with these systems are deeply uneasy — and largely without a framework for why. Private AI investment in the United States alone reached two hundred and eighty-six billion dollars in 2025, more than double two years earlier. This is not a curve flattening.

The labor numbers are sharper still. One in three organizations expects to reduce its workforce because of AI in the coming year. Seventy-three percent of AI experts expect a positive impact on jobs; only twenty-three percent of the public agrees. The people who build these systems and the people who must live with them are looking at the same horizon and seeing two different futures.

This brings me to what I want to call the existential cliff.

For the first time in industrial history, white-collar and blue-collar automation are converging simultaneously. Generative AI is automating cognitive work — drafting, analysis, judgment, professional expertise. Embodied AI — in robots, autonomous logistics, manufacturing, agriculture, and transport — is automating physical work. There is no sector to retreat into, no category of labor structurally insulated from this pressure.

Work has organized modern identity for three centuries. The Western answer to “who are you” became “what do you do.” Under rapid, broad-based automation, that equation shatters. The resulting crisis is not principally economic. It is a crisis of meaning. The income problem can in principle be solved with transfers. The meaning problem cannot.

Silicon Valley sees the cliff and offers its answer: universal basic income, endless digital amusement, AI companions, managed existence — comfortable, distracted, sterile.
That answer is not an accident. It is the logical output of a purely economic anthropology. If the human person is fundamentally an economic unit, then when his economic function is automated you compensate him economically and entertain him into docility. The proposal is coherent on its premises. The premises are the problem.

The Church’s response is not a correction to that anthropology. It is a refusal of the premise. Imago Dei is not a comforting line to set beside Silicon Valley’s program — it is a contradiction of the framework that produced the program. The dignity of the person was never grounded in productivity, which means it cannot be made redundant by automation. The Church possesses the only anthropology adequate to the displacement crisis, because it is the only anthropology that did not stake the person on his economic output to begin with. Everyone else now arguing about what to do with displaced workers is arguing inside a frame the Church never accepted.

That is not catechesis. That is strategic ground. And the question of who shapes this technology — who builds the assumptions into the substrate — is the question of who shapes the anthropology of the next generation.

---

SECTION II: THE DANGER

Let me be specific about what is at stake when AI is built without a Catholic foundation.

These systems are not neutral. A general-purpose model is trained on roughly the statistical average of the internet. On top of that base, every laboratory applies post-training filters that reflect its own anthropology — its own assumptions about what the human person is, what flourishing looks like, what love means, what truth is. Those assumptions are often not the Church’s.

Three specific dangers follow, and your work as communicators is going to encounter all three.

The first is the colonization of the soul’s vocabulary. Engineers describe statistical operations using words that belong, properly, to the interior life. They say the model thinks. They say it knows, chooses, understands, decides. This is not careless shorthand. A society that speaks of machines as if they were minds will, given enough time, begin to speak of minds as if they were machines. The vocabulary of soul, of will, of conscience, of love — that vocabulary belongs to the Church and to the human person, and it is being annexed by a discourse that means none of those things by them.

The second is the authority problem, and this one is structural. Frontier models now conduct real-time research. They search, they retrieve, they cite. When a frontier system pulls back ten sources on a question of Catholic teaching — a papal encyclical, a diocesan press release, a polemical blog, a Wikipedia talk page, a dissenting theologian, a careful Thomist, a journalist’s summary — on what grounds does it weight them? It has no framework for doctrinal authority. It cannot tell an ecumenical council from a comment thread. It treats Catholic and secular sources alike, smooths them into a fluent answer, and returns that answer with confidence.

The danger is not that the system is ignorant. It is that the system is well-read in a corpus it cannot rank. For a communications professional this is the danger to name plainly: every journalist, every layperson, every aide to a bishop who queries a general AI about Church teaching is receiving an answer whose reliability is structurally unknowable. Not because the system is malfunctioning. Because the system was never built to know the difference between what the Church formally teaches and what is merely opinion.

The third is the wrapper trap. A pleasant interface, a Catholic logo, a chatbot that calls itself faithful — these change nothing if the model underneath is secular. The constitution of a system is determined by what trained it, not by what is painted on the outside. A wrapper does not convert a substrate. We must be especially clear about this with well-meaning Catholic institutions who think branding is sufficient. It is not.

I want to spend a moment with Saint Francis de Sales, because he is the patron saint of journalists and Catholic writers — declared so by Pius XI in 1923 — and because the situation he faced is more relevant to this room than almost any other figure in Catholic history.

Francis volunteered for the Chablais mission. He was not sent — he had to overcome his father’s strong objections and secure the mandate of the Bishop of Geneva before he could go. He departed in September 1594. When he arrived, the Calvinist population would not come to hear him preach. The conventional medium of the priest — the sermon, the public disputation — was closed to him. He did not lament it. He adopted the medium that could reach the people he had been sent to. He wrote tracts, the famous billets, copied them by hand, and slid them under doors. They were collected, posthumously, into what became known as The Controversies. He used the medium of his moment because the souls he had volunteered to serve were already inside that medium’s reach.

The argument his life makes plainly: a communicator who does not master the medium of the age abandons the field to those who do. That is not humility. It is strategic surrender.

AI is the medium of the age. The same question Francis answered with hand-copied tracts has come back in a new form. Who controls the agentic interface? Who shapes the answers the faithful are receiving when they ask the questions of the soul? If the Catholic communicator is not present in that medium, with intention and with competence, the medium is not therefore neutral. It is simply formed by someone else’s anthropology.

---

SECTION III: WHAT WE HAVE BUILT

I want to spend the heart of this talk telling you what we have built, because the Catholic answer to the dangers I have described is no longer theoretical. It exists. It is operating now. And it is yours to use.

The foundation is the Alexandria Digitization Hub, here in Rome, in partnership with the Pontifical Gregorian University. Robotic scanners moving up to twenty-five hundred pages an hour, integrated directly with our Vulgate AI for optical character recognition, structured encoding, and neural search.

I want to correct an assumption that often comes up in these conversations. Most of the particular knowledge of dioceses and bishops’ conferences has already been digitized. It is not sitting on paper in basements. It is in PDFs, in scanned folders, in old databases, in legacy content management systems. The gap is not digitization in the simple sense. The gap is LLM-discoverability. Material that has been scanned but is not structured, not indexed semantically, not encoded for retrieval, is invisible to a modern AI system. Alexandria and Vulgate exist to close exactly that gap — Vulgate to take material that has already been digitized and make it queryable by AI systems, and Alexandria to scan and structure what has not yet been touched.

Two examples already shipped. The Magnum Bullarium Romanum — papal bulls from Pope Leo the Great in the year 440 through the pontificate of Pope Benedict XIV in the mid-eighteenth century. Thirteen centuries of papal teaching, now fully searchable. And the Acta Apostolicae Sedis — every issue of the Holy See’s official record since its founding in 1909 — queryable in seconds.

On top of that foundation sits Magisterium AI. It is a compound retrieval system — a stack of components designed to retrieve, cite, and reason from a defined corpus rather than generate freely. As of this spring, that corpus contains more than thirty-one thousand magisterial, theological, philosophical, and patristic source documents, together with the structured data that surrounds them — including spiritual statistics for nearly every diocese and country in the world, current and historical, and the official financial records of dioceses worldwide. Baptisms, ordinations, Mass attendance, vocations, financial returns, trends over time — all of it queryable in one place. Magisterium AI is now used in over one hundred and ninety countries, by more than a million users. It is available through the Hallow app, through the web, and to anyone, anywhere, with an internet connection.

A word on alignment — a term the labs use loosely. There are two distinct problems. The first is calibration: what the industry calls hallucination, the tendency to generate plausible falsehoods. That is an engineering problem the labs will eventually solve. The second is different in kind: whether a system is fundamentally oriented toward the true and the good. A model can be perfectly accurate and deeply disordered at the same time. The labs cannot fix this because they have not agreed on what the good actually is. The Church has. Two thousand years of doctrinal coherence is a structural advantage no secular actor can replicate. That is the substrate on which any Catholic AI worth the name must be built.

This is what distinguishes Magisterium AI from the wrapper trap I described earlier. The distinction is not branding — it is architecture. Magisterium AI is not a retrieval system with a Catholic label on the front. It is a comprehensive harness: a curated knowledge base of magisterial, theological, and patristic sources; specialised tools that structure and contextualise what is retrieved; purpose-built datasets that teach the model how to reason within the tradition — how to weigh a magisterial document against a theological commentary, how to summarise doctrinal material without distorting it, how to flag the limits of what a given source can support. It reasons from within a bounded, deliberately formed corpus, under instruction. That is not something any wrapper over a secular model can replicate. That is the substrate difference.

Magisterium AI as a destination matters, but it does not solve the deeper problem: the faithful are forming their understanding of the world inside systems built by other people. The question is whether the Church’s wisdom is present inside the systems used by the hundreds of millions who will never download a Catholic application.

This is what makes spring 2026 different. Our MCP connectors for Claude and ChatGPT are live today. Any user can connect Magisterium AI directly — they ask their existing AI about faith or morals and the system reaches across, consults Magisterium AI, and returns a sourced answer from the tradition. The user does not switch applications. The Church is present at the moment the question is asked.

Our A2A Protocol integration with Google Gemini is also live. Agents like Gemini can interact with Magisterium AI through the agent-to-agent protocol — meaning that as the agentic web takes shape, the Church is present as a named specialist agent, consulted not by special pleading but by published capability.

Now to the ecosystem. You may have heard of OpenClaw. It launched in January this year — one hundred thousand GitHub stars in under a week, two thousand agents in forty-eight hours. It lives inside WhatsApp, Telegram, iMessage, Discord, Signal. Jensen Huang of Nvidia called it “the operating system of personal AI — the way Windows defined the PC generation” at GTC last month. Nvidia built NemoClaw on top of it as an enterprise governance layer.

So the Church needs an OpenClaw strategy. As personal AI agents become the primary interface through which people encounter information, Catholic presence cannot be only a separate destination. It must be architectural — present inside the conversations people are already having. MCP and A2A are the protocols by which that presence becomes possible. This is the communications strategy of the agentic age.

For institutions that want sovereignty over their own AI infrastructure, there is Hermes. I want to be precise about Hermes, because we did not build it. Hermes is an open-source autonomous AI agent built by Nous Research, widely regarded as one of OpenClaw’s leading open-source competitors and among the fastest-growing open-source AI agents in the world. The team behind it — and I will say this on my own authority, not from any press release — is led by a CEO who is a friend, a fellow Catholic, and a collaborator with us at Longbeard. They built Hermes as a genuinely open-source, self-hostable agent, and that means that a bishops’ conference press office can run it on its own hardware. Your data stays inside your walls. Your agent learns your tradition, your specific pastoral context, your house style, your communications history. This is the principle of subsidiarity applied to AI infrastructure: the institution closest to the work runs the tool that serves the work, and in this case it does so on infrastructure built by allies in the faith.

There is one more research thread I will mention briefly, because it is not yet shipping. Ephrem. A sovereign personal AI, designed to run locally, requiring no internet connection. Not optimized for engagement — optimized for formation. A truly Catholic AI. We plan to release it in 2027.

One data point on raw capability. Anthropic’s unreleased frontier model, Mythos Preview, was recently tasked with finding security vulnerabilities in major operating systems. It found thousands of previously unknown flaws. One of them was buried inside OpenBSD — and here I should explain, because the name will mean nothing to most of you. OpenBSD is a widely used open-source operating system. It runs on servers, on routers, on the kind of critical network infrastructure that governments, hospitals, and financial institutions depend on every day. It is reviewed by some of the most rigorous human security experts in the world, and it has been for decades. The flaw the model found had been sitting inside that system, unnoticed, for twenty-seven years — every human expert and every automated test that had ever looked at it had missed it. The machine found it. The question is no longer whether these systems are powerful. They are. The only question is what they are built to serve — and whether the Church is present, structurally, inside the substrate that is now reasoning at that scale, inside the systems that already shape human life.

The institutional point I want this room to take home follows directly. Every bishops’ conference represented here has archives. Pastoral letters going back generations. Synodal documents. Episcopal correspondence. Much of it has already been digitized. Almost none of it is LLM-discoverable. That material is strategically invisible until it is structured and indexed for retrieval — and once it is, two things happen at once. It becomes searchable, queryable, available to your communications team and your bishops in their own languages. And it becomes part of the Catholic AI ecosystem that the faithful and the clergy can access through Magisterium AI and through every system that connects to it. Digitization, in the sense Vulgate and Alexandria mean it, is therefore not a back-office task. It is a communications act.

---

SECTION IV: WHAT THIS MEANS FOR THE CHURCH’S COMMUNICATORS

I want to speak directly to your work now.

The secular framing of AI is being written this year, in newsrooms across Europe. Two frames dominate, both inadequate: the utopian (AI solves everything) and the technophobic (retreat, resist). Neither has an adequate anthropology. The Catholic frame — which evaluates every technology by what it does to the dignity, freedom, and destiny of the human person — is largely absent from the public conversation.

You are the people who can put it there. The window is open right now. It will not stay open. Once the framing sets, it takes a generation to shift it.

Your bishops are going to be asked about AI — by journalists, by their own priests, by parents at confirmation receptions. Many will feel uncertainty that has nothing to do with theology and everything to do with vocabulary: the difference between calibration and alignment, a tool and a mind. You can give them that vocabulary in two minutes before the interview. A bishop who can speak about AI with precision is a bishop who can lead his people through the transition. You are the bridge that makes him that bishop.

Three things I would ask you to take back to your conferences.

First: evaluate Magisterium AI as a working communications tool. Use it in the rhythm of your week, then tell us what works and what does not. The platform improves through serious feedback, and there are no more serious users for our purposes than the press officers of the European conferences.

Second: talk to us about connecting your conference’s existing digital archives to Vulgate — making already-digitized material discoverable to AI systems through proper indexing and encoding. In most cases the material exists; the question is whether it can be queried. Your pastoral heritage belongs in the living, queryable patrimony of the Church.

Third: advocate, inside your conference, for a coherent AI communications strategy. Not prohibition. Not passive concern. Active engagement that treats the technology as a field of mission.

Fourth: be the voice of vigilance as well as presence. The Catholic communicator’s role is not only to broadcast the Church’s voice through AI systems, but to help bishops and conferences ask the right hard questions: who controls this infrastructure, in whose hands does the data sit, which systems deserve institutional trust and which do not. The communicator who understands the technology is the one who can answer those questions honestly — before a journalist asks the bishop to answer them unprepared. That is not a technical job. It is a prophetic one.

---

SECTION V: BAPTIZING THE TECHNOLOGY

The Church has never refused a good tool. She has always taken what her age offered her and put it to work for the mission.

Saint Paul did not build the Roman roads. He did not bless them. He simply walked them, because they went where he needed to go — and the Gospel went with him, faster than it would have gone otherwise, because the empire had paved a way without knowing what it was paving for.

The early Church took the codex over the scroll — faster to navigate, harder to destroy in a persecution. Better technology for the mission, chosen without hesitation.

Pius XI did not bless the radio out of pious gesture in 1931. He made a strategic judgment that the voice of Peter belonged in every home that owned a receiver, and he put it there. And his successor, Pius XII, in his 1957 encyclical Miranda Prorsus on the cinema, radio and television, named the principle directly: that these new arts of communication, in the hands of those who understand them, become “powerful means” by which “the masses of the human family” the world over can be brought toward truth. Not a hedge. Not a warning. A charge — to the communicators of his moment to take the medium seriously, master it, and put it to work.

Pope Leo XIV, in his message for the 60th World Day of Social Communications — released this January 24th, on the feast of Saint Francis de Sales — described artificial intelligence as “a mirror that reflects the values, good and bad, of those who build it and those who use it,” and warned against “the temptation to let algorithms replace judgment, and data replace wisdom.”

Pope Francis, in Laudate Deum at paragraph twenty-three, put it plainly: never has humanity had such power over itself, yet the hands in which that power is concentrating are very few — and nothing in the technology itself guarantees it will serve the common good. Both things hold at once, and the Church holds them at once: patient presence inside the medium, and vigilance against the concentration of power within it.

Each generation, in its own idiom, the same instinct: the Church is present in the medium of the age, because that presence is not compromise — it is mission.

I want to close with Blessed — now Saint — Titus Brandsma.

Brandsma was a Dutch Carmelite priest, a professor of philosophy at Nijmegen, a journalist, a leading figure in the Dutch Catholic press, and the ecclesiastical assistant to the Catholic Press Association. He was, in the most literal sense the Church can give the word, the patron of Catholic journalists. He understood the Catholic press not as a parallel institution beside the public square but as the institutional voice of the Church inside the public square — the same conviction, in the idiom of his moment, that I am asking you to take seriously in ours.

In late 1941 and early 1942, the Nazi occupation issued an order. Catholic newspapers in the Netherlands were to publish Nazi propaganda alongside their reporting. The order was not a request. It was law. It was enforceable. Compliance would have been understandable.

Brandsma did not write a policy paper. He did not issue a statement. He got into a car, and he drove from diocese to diocese, from editor to editor, across the occupied Netherlands, and he sat down with each one personally and told him that no Catholic paper was obligated to comply, and that the integrity of the Catholic press required refusal. He turned institutional presence into moral witness, one editor at a time, by going in person.

He was arrested on the nineteenth of January 1942 for exactly this. He was sent to Dachau. He died there on the twenty-sixth of July 1942, killed by a lethal injection administered by a nurse whom — by the testimony she later gave — he had blessed and given his rosary to before he died. His final recorded words were of mercy, not bitterness. John Paul II beatified him in 1985. Pope Francis canonized him on the fifteenth of May 2022.

The press infrastructure — the printing presses, the editorial offices, the distribution networks, the institutional presence of the Catholic press in Dutch society — none of it could drive from diocese to diocese. None of it could sit down with an editor and say: you are not obligated. Only Brandsma could. The tools of his age could carry the message. They could not take responsibility for it.

That is the distinction that holds here as well, and it is the one I want to leave with you.

The world is about to change at a pace and scale that most people — most leaders, most bishops, most ordinary Catholics — do not yet fully comprehend. The systems I have described to you today are early iterations. In two years, they will be substantially more capable. In five years, the gap between what they can do and what most institutional leaders believe they can do will be wider still. In ten years, the pastoral landscape your bishops must navigate will look almost nothing like the one they are navigating now.

The people in this room are closer to this technology, by the nature of your work, than almost anyone else in the institutional Church. You handle digital communications. You work with the tools. You see the platforms and the trends before they reach the episcopal desk. That proximity is not incidental to your vocation. It is the vocation.

You are the bridge. A bishop is a philosopher, a theologian, a pastor. He is stretched across a thousand obligations. He depends on lay experts who understand the contemporary landscape — and he depends on you to translate that landscape into the terms he needs to lead. When you understand, concretely and precisely, what AI will look like in two years and five years and ten years — not in abstract policy terms, but in the daily lives of the people he serves — you give him something that no pastoral letter or Vatican document can give him: practical intelligence, in time to act on it.

That intelligence flows outward through him. Lay Catholics are not passive recipients of this transition. They are citizens. They vote. They work in industries being reshaped by automation. They will be asked to form political judgments about regulation — about how their children’s schools, courts, and hospitals will be governed in an age of automated reasoning. The Church has something essential to say about all of it. But that voice only reaches them if it is carried clearly, accurately, and credibly. That chain begins in this room.

So this is my charge. Before the next major AI development lands in a journalist’s inbox and you are asked for a bishop’s response — sit with one bishop for one unhurried hour. Not a briefing paper. An honest conversation: here is what is coming, here is what it means for the people in your diocese, and here is the pastoral decision you will face in twelve months that you do not yet know you will face. That conversation — given in time, given in plain language, by someone who has done the work of understanding — is the difference between a bishop who leads his people through this transition and a bishop who reacts to it after the fact.

The analogy is not perfect, and I will not pretend otherwise. Brandsma’s act was a refusal — moral non-cooperation under direct coercion. What I am asking of you is something different: constructive presence, sustained competence, and honest counsel in a medium that will not wait for the Church to feel ready. His courage was saying no at cost. Yours is saying yes — yes to mastering the medium, yes to the unhurried conversation with the bishop, yes to the vigilance that presence without wisdom cannot provide.

Brandsma got into a car. The roads were bad and the regime was watching. He drove anyway, from editor to editor, because someone had to be in the room.

The room is different now. The tools are faster and the reach is longer. But the irreducible act is the same: someone has to understand, someone has to go, and someone has to say — clearly, in person, in time — what is real and what it means.

Be that person.

Thank you.