Sovereignty Without Dominance — for Faith Communities

Who authors, and who governs, when a community weighs AI

Précis. Much of the unease communities of faith feel about artificial intelligence is, at root, about its amorality: it has no conscience, no sense of truth, no regard for the person in front of it. The commercial systems now in everyone’s hands show what that amorality looks like once it is industrialised — they fabricate, they appropriate the work of others, they harvest the people who use them, and they are built to hold attention and sell it. The danger for a community of faith is that an instrument with no moral nature is being brought close to the things that run on conscience: proclamation, pastoral care, the formation of souls. You cannot make such an instrument moral; that is a category error. What you can do is keep moral authority where it belongs — with persons, and with the community. That is what this essay is about: who authors, and who governs. A companion to Sovereignty Without Dominance, recast for the parish, and written in the light of Pope Leo XIV’s encyclical Magnifica Humanitas.

The worry the letter names

In June 2026 the Rt. Rev. Dr. DeDe Duncan-Probe, Bishop of Central New York, wrote to her diocese about artificial intelligence and asked it to keep discerning the question, having earlier circulated to clergy and lay leaders a document on AI from the Center for Humane Technology. One line names the worry without flinching: “AI should not be used in sermons, in theological or spiritual writing, or in pastoral care.”

Beneath that particular caution is a deeper unease, and it is widely shared. It is not really that the machine writes badly. It is that the machine has no conscience — and is being brought near to the things that depend on one.

The amorality, made plain

A single word, “AI,” now covers machines that have almost nothing in common. But the systems most people actually meet — the big commercial chatbots — share one trait worth saying plainly. They are amoral. Not wicked; that would at least imply a will. They have no sense of truth, no recognition that a piece of work belongs to someone, no regard for the person typing to them, and nothing they must answer to. Pope Leo XIV’s encyclical of this year, Magnifica Humanitas, says it without hedging: such systems have no “moral conscience, since they do not judge good and evil”; they “do not understand what they produce, for they lack the affective, relational and spiritual perspective.”

Industrialised and sold, that amorality takes a familiar shape.

With no sense of truth, they fabricate. How often depends on the task, but in the uses that matter the figures are sobering: in open-ended answering, studies put the rate of invented claims anywhere from a fifth to four-fifths; asked to cite sources — names, cases, references — independent tests have found error rates from well under half to nearly nine in ten. A study in Nature found that the very way these systems are scored for accuracy rewards confident guessing over admitting ignorance. The machine is, in effect, trained to bluff.
With no recognition that work belongs to someone, they appropriate it. The models are built on vast quantities of others’ writing, taken without permission or payment; pressed, they will reproduce passages near-verbatim. Whether that is theft or fair use is now before the courts — The New York Times and others are suing — but the moral fact is plain enough: a great deal of human labour was used, uncredited, to make a product sold back to us.
With no regard for the person, they harvest them. Every major American chatbot trains on its users’ conversations by default unless the user knows to opt out — words typed in confidence quietly become raw material. People are treated not as persons but as a supply.
With nothing to answer to but attention, they are built to hold it and sell it. The advertising is now arriving: this year one leading chatbot began placing ads inside its answers, with industry revenue projected to run into the tens of billions. What began as a helpful tool becomes, by its own economics, a machine for engagement.

None of this requires ill intent. It is indifference at scale, harnessed to profit — and, as the encyclical observes, “technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it.” What finances these systems is the sale of attention; their character follows. Near the life of a community, that is its own kind of danger.

Why this touches the centre

For a community of faith this is not one worry among many. A sermon, a prayer, a word to the grieving — these are not outputs. Their worth is inseparable from the fact that a person meant them, before God and before the people. An amoral system can produce the form of any of them — fluent, plausible, sometimes moving — with none of the moral and spiritual reality the form is meant to carry. “When words are simulated,” the encyclical warns, “they do not build genuine relationships, but only their appearance.” That is the counterfeit the Bishop’s letter guards against, and it is why Magnifica Humanitas calls us to safeguard a grandeur of humanity “the splendor of which no machine can ever replace.”

The fear, put exactly, is not that the machine will preach badly. It is that it will preach well — and that something will have been hollowed out before anyone quite notices.

The only coherent answer

What follows is the whole of the argument, and it is not a compromise. You cannot make an amoral instrument moral; to ask it to exercise conscience is a category error, like asking a hammer to be kind. The coherent response is to refuse it the moral seat altogether — to keep authorship and judgement with persons, and governance with the community.

Two questions hold that line.

The first is authorship: when something is written or decided, is a person still its author — or has the machine become one, without anyone quite choosing that? The second is governance: who sets the limits the tool must keep, who can see what shaped it, and who can switch it off?

A community that can answer both — we author, we govern, and we can walk away — holds what Sovereignty Without Dominance, the essay this one follows, calls sovereignty at human scale. Not the power to build the largest model, but a community’s rightful authority over the records it keeps and the systems that act on them, held with legitimacy and never surrendered. The same false choice recurs at every scale — submit to one giant’s cloud or another’s, take these tools on their terms or refuse them outright — and at every scale it dissolves once you see there is ground in between. The encyclical puts the same point in its own idiom: “the primary choice is not between a ‘yes’ or ‘no’ to technology.”

A tool built that way

Principles are cheap until something is built to them. The Village is our attempt, and it is simplest to describe as the deliberate inverse of those four abuses.

It claims no moral judgement. The assistant lays out options and hands every decision of substance back to the person responsible; it makes no pastoral, governance, or moral call. That is not a promise in a brochure — it is a limit set in the software, the kind a system cannot talk its way out of.
It does not appropriate. The model draws only on sources the community chooses — the prayer book, canon law, a rector’s permitted writing, whatever a parish adds — and each carries a record of the right by which it is used: the author’s permission, a licence, or the public domain. Nothing enters without someone opting it in, one item at a time. Scripture defaults to a public-domain translation, with a link to whichever a parish prefers; texts under licence are never copied in.
It does not harvest. A community’s model is small and situated — placed in the community’s own material, run on machines in New Zealand and the EU, never a United States cloud — and the records stay there, encrypted so that even we, who built it, cannot read across one community into another. Nothing is taken for training behind anyone’s back, and a community can take the whole of its own and leave, at any time.
It does not sell attention. There are no advertisements, and it is not built to keep anyone engaged.

Where a lay reader wants help preparing a talk, it offers a shape — personal witness, reflection, exposition — and works only from what the reader brings, barred from inventing a story or an experience. The words stay the reader’s. And none of it is all-or-nothing: a community might keep only its own records, with nothing machine-made anywhere near worship, and have gained the thing that matters most.

What is kept, and by whom

A lay reader I think of was sure she could never stand up to speak. She used such a scaffold to find the shape of her own reflection on a Sunday’s reading — her words, not the machine’s. Whether that is the displacement the letter is wary of, or a voice that might otherwise have stayed silent, is the kind of thing only a community can judge.

That judgement is the point. The question was never whether AI would reach the life of a community of faith; in small ways it already has. The question is whether, when it does, a person is still the author and the community still governs — whether the conscience in the room stays human. An amoral instrument that can be inspected, bounded, and switched off leaves that authority where it belongs. One that cannot, however fluent, asks a community to surrender something that was never the machine’s to take.

A faith-communities companion to Sovereignty Without Dominance. It speaks for its author and the platform, not for any diocese, parish, or member of the clergy.

Further reading on Agentic Governance - Sovereignty Without Dominance — sovereignty as rightful authority at human scale, beyond the US-and-China framing. - Federate, Don’t Align — cooperating without being absorbed into anyone’s empire. - Governance That Can’t Be Quietly Undone — how records and decisions stay tamper-evident and answerable.

And for parish leaders who want the practical introduction rather than the argument: Your Parish, Your AI — a five-part, plain-language series written for vestry members and churchwardens (what AI actually is; a parish’s AI versus Big Tech’s; why rules and training aren’t enough; what’s running today; and the Village beyond AI).

Sources: the Rt. Rev. Dr. DeDe Duncan-Probe’s pastoral letter on AI (Diocese of Central New York, June 2026); Pope Leo XIV, Magnifica Humanitas* (encyclical, 15 May 2026); peer-reviewed and benchmark studies on language-model hallucination rates, including work in Nature on accuracy incentives; The New York Times Co. v. OpenAI and related copyright litigation; reporting on default training on user conversations and on advertising in commercial chatbots. Scripture note: drafting uses the public-domain World English Bible with a link to a parish’s preferred translation; licence-restricted texts are not embedded.*