Should religious AI chatbots be treated differently from all others?

Yes, no, how, and above all: why?

(please DO read this to know what I write here, why and how, and how to support it. Thanks!)

Magisterium AI is a chatbot (free for 10 queries) that’s presented as “the world’s leading Catholic AI”. Its developers obviously present it as something unconditionally good, because it’s about True Faith. Its opponents call for its destruction, because it’s about True Faith.

What is truth here, about Magisterium AI or any other religious chatbot for that matter? That’s a question everybody should consider and, unlike Pilate, here you can stay for an answer.

With Magisterium AI, what’s the basis? For which charge, exactly? And what should the penalty be? (Image source: John 18:38 in Bible Hub).

“Instant, trustworthy answers” (remember ,”instant”)

Magisterium AI offers “instant, trustworthy answers rooted in Catholic tradition to explore faith with clarity and depth, get accurate answers to specific questions about faith and morals and clarify complex doctrines (like the Trinity or Sacraments) faithfully”.

Longbeard, the company that develops this chatbot, says it trained it exclusively on the world’s largest database of Church documents: almost 25,000 magisterial documents and (in Scholarly Mode) over 4,333 Catholic theological and philosophical works from St. Thomas Aquinas, St. Augustine, Fathers of the Church, etc.

This, says Longbeard, makes Magisterium AI avoid hallucinations and provide answers “grounded only in authentic Catholic teaching and tradition with specific citations to every assertion it makes, minimizing the risk of doctrinal error and giving confidence that the information aligns with the mind of the Church”.

Should Magisterium AI be deleted?

In January 2026 Marc Barnes, editor of New Polity, published his reasons why Longbeard should delete Magisterium AI. Some of those reasons are purely philosophical or theological. Barnes argues that conversations about faith should only happen between humans, but no AI chatbots can really converse about anything and that, regardless of the quality of the sources of Magisterium AI, it cannot use them properly anyway because “faith should never be probabilistically specified. A dice-roll is still a dice-roll even when it always lands on precisely the “6” you want... The Holy Faith is never simply “the right words in the right order,” but a Credo, an “I believe in”“ that no machine could ever say.

One part of the answer by Harvey Sanders, CEO of Longbeard, isn’t enough for me to agree with Barnes, but really rubs me the wrong way: “I leave the deep metaphysical distinctions to the scholars who are far more learned than I. I am a builder.”

Sorry, but... isn’t that the same, stale “Move Fast and Break Things” Big Tech religion that had already made a mess of society and political discourse before AI? “Trust we who can code because we can code and code we will, will ye, nill ye”? Thanks, but no thanks, on principle.

Sanders, certainly in good faith, also insists that Magisterium AI is “just a tool”. Sorry again, but no, no technology is just a “neutral” tool, especially if it’s digital. One year ago, I explained that current digital social media are toxic because they just repeat the same architectural and urban design errors that make physical cities mentally toxic.

This year, Barnes said something very similar: the very design of any machine, including chatbots, presumes and contributes to mold a certain kind of human being and society, regardless of how much its “builders” realize or want it. Since the philosophy behind chatbots is not really compatible with the Catholic one, Barnes says, they could never be neutral “tools” completely adaptable to Catholic ends. I tend to agree, but even this wouldn’t be enough for me to call for an outright deletion of Magisterium AI.

The other two main arguments in Sanders’ defence of Magisterium AI are one implicit (”every digital technology is great”) the other very explicit, that is FOMO (”Fear Of Missing Out”) and TINA (”There Is No Alternative”), which is well represented by these three almost literal quotes from his reply:

1. We are already crossing the river that separates the world before AI and the world with AI. The question is: who will write the code that governs the other side?

2. If we “delete” our presence in this sphere, we don’t stop the sphere from existing; we simply ensure it remains godless.

3. We must have the courage to baptize the tool, not bury it.

Now, regardless of what people who will lose big money when (NOT “if”!!!) the AI bubble pops desperately claim, what we call AI today is NOT AI and above all is not unavoidable. Not at all. To see why, you can read what a veteran tech journalist defined “one of the best articles on AI you’ll read this year”.

There are plenty of good things we should do as soon as possible with a saner version of AI, starting with agriculture. But accepting AI everywhere for everything just because it’s here is the same mistake already made with IoT, NFTs, the metaverse and soon, if we keep being dumb,humanoid robots.

But even if AI everywhere was truly unavoidable, so what? If some version of some technology or practice is objectively unavoidable but also bad or dangerous... shouldn’t Catholics and everybody else still strive to limit as much as possible its usage in every form? If “unavoidable” were a sufficient reason to make one’s own version of something, the Catholic solutions to prostitution and smoking could be Catholic brothels and cigarette factories.

Those are paradoxical examples of why “just make a Catholic version of something and it will be fine” isn’t a great strategy, but Barnes makes the same point much more seriously.

“Let’s make a Catholic version”, he notes, is the same thing lay people have been always encouraged to do with any other media before chatbots: newspaper, radio, television, computers, the Web... but in the end the only places where Catholicism grew instead of just slowing down its retreat are those “with the least and tardiest access to each new technological instrument of mass communication” but bigger size and strength of families, or stronger missionary presence. Here is another way Barnes puts it: “Places that have good things don’t need technologies which produce online versions of good things.” That is, social innovation is always much more powerful than technology alone. Even when the task is to make places “not remain godless”.

My extra reason to dislike the defense of Magisterium AI

Sanders passionately rejects any conclusion that “we must delete this technology to preserve our humanity [because doing so] confuses the medium with the mission [and] the mechanics of language with the intimacy of conversation..., [Q]uerying in natural language is [not] the same as conversing”, and Magisterium AI just offers “intellectual clarity, not sacramental grace”.

What Sanders does there, again surely in good faith, is the same mistake that Free / Open Source traditional advocates have made for decades, and I already denounced 20 years ago in the Seven things we’re tired of hearing from software hackers, and the Free Software Manifesto for the rest of us:

the belief, absolutely detached from reality, that since software “builders” were born with the capability to write and use software in a certain way, everybody else will eventually do the same and love it, will he, nill he.

I don’t doubt that Sanders never confuses clarity with grace, or querying with conversing. I am just equally sure, because that’s what I’ve seen happening every day with Gnu/Linux, Web 2.0, social media, AI bots and every other software since the mid-1990s, that 90% of real people are not like him, and WILL indeed confuse themselves. Especially after having been dumbed down by 20 years of social media designed with the very purpose of isolating and dumbing everybody down.

This is not judgment or criticism of anybody. This is just an acknowledgment that there are millions of different types of human brains, all equally worthy, but those wired to react to software exactly as Sanders pictures are a tiny, tiny minority.

Sanders’ other statement that Longbeard is not replacing authority, just amplifying access to it thanks to the “instant, trustworthy answers” of Magisterium AI is just as weak and worrying: “Trust us, human brains and souls will only benefit from always-on media and super-fast information with instantaneously amplified feedback and amplified, worldwide exposure!”... is exactly what Zuckerberg, Altman and Co have been pushing on the world for two decades now, and we all know how it worked.

Screenshot from goodreads.com.

So. should Magisterium AI be deleted? No, it should just...

In spite of everything I wrote, I don’t think there is a need to delete Magisterium AI as Barnes wants. I think Sanders himself unwillingly showed what to do in his answer, when he wrote that:

1. When a student queries Magisterium AI, they are not seeking “communion” with the software any more than a scholar seeks “communion” with the card catalog or the index of the Summa Theologiae. They are seeking access to the patrimony of the Church

2. Magisterium AI acts less like an author and more like a paralegal. It locates the precedent, summarizes the specific text, and places the document in front of you.

Exactly! Thanks Mr. Sanders! Paralegals are legal professionals who do legal research or drafting legal documents only for, and under the direct supervision of a lawyer!, not for her clients. Nobody (OK, nobody in his mind...) who needs legal help would hire a paralegal instead of a lawyer, even if paralegals were allowed to work as lawyers. Paralegals cannot have clients.

On top of that, anybody who’s really, deeply competent in some topic will tell you something like “I regularly see AI chatbots making mistakes about that topic; therefore, I exclusively use them for that topic, not for any other purpose where I couldn’t spot and avoid errors by myself”.

Add that fact to the limits of paralegals, and the solution is simple: instead of deleting Magisterium AI,

1. immediately change its name to something more sincere like “Magisterium Assistant”. In 2026, “AI” is already attached to way too many product names to look anything but FOMO marketing, and once the bubble pops it will become a liability anyway

2. only, only license the “Magisterium Assistant” to priests, or people like theology professors with at least five years of experience and the knowledge to properly evaluate Barnes philosophical objections

In other words, it’s practically certain that the general public will misuse and abuse Magisterium AI just like it has done with ChatGPT and any other “know-it-all” chatbot since 2022. Therefore, there is no reason why Magisterium AI (or any similar service for any other religion, of course) should not be treated as I already recommended two years ago for every AI chatbot like ChatGPT: just block universal access to it.

While I’m at this, everything Sanders says is for me further proof that the future of AI - that is of AI applications that make sense and don’t bankrupt a whole planet or suck up all the energy it needs, is not AGI, but SMALL, single-purpose AIs. Which hey, is just what China is doing, unlike the US.

Final note: the next piece will also be about Catholics vs (open) software in the 21st century and why everybody should care, so stay tuned!