EkklisiaAI for the ΑΩ

English · Διαβάστε στα ελληνικά →

Churches are starting to govern AI, not just use it

19 November 2026 · 10 min read

The question is shifting — from “how can we use AI?” to “what rules do we set for it?”

For two years the church conversation about AI has mostly been about use: which tool drafts the newsletter, which one transcribes the sermon, which one is free. This week, on two continents, it became about government. On Tuesday 17 November, Argentina's evangelical churches opened an AI observatory. On Wednesday 18 November, the US Catholic bishops voted AI rules into their investment guidelines. Neither body was announcing a new tool. Both were setting up a way to keep answering the question — and that is the shift worth noticing.

Argentina: an observatory, with children first

On the afternoon of Tuesday 17 November, the Christian Alliance of Evangelical Churches of the Argentine Republic (ACIERA), the national umbrella body for Argentina's evangelical churches, launched an Artificial Intelligence Observatory. Pastors and church leaders from Entre Ríos, Santa Fe, Chaco, Buenos Aires province and the capital attended, in person and online. Its motto: “On the altar of productivity we cannot sacrifice human dignity.”

ACIERA's president, pastor Christian Hooft, acknowledged the benefits first — “indisputable scientific and medical advances in terms of speed or efficiency” — and then named the price: “as nothing is free, in that productivity we give something up.” The church, he said, had a responsibility “not to leave this duty of care only to the State.”

His sharpest concern was children. “We are transforming the social life of children as in an experiment,” he said, and the observatory wants to study which legal frameworks to work on “to avoid harm to their capacity for discernment.” He described the voice he wants the observatory to have as “prophetic,” and rejected an “apocalyptic” one.

The observatory has a director, Ayelén Caffarena, a lawyer with a master's degree in government, and a small team: Viviana Barrón, a doctor of social sciences and rector of a Baptist theological seminary; Darío Acosta, a specialist in communication, education and new technology; and María Florencia De Liguori, a psychologist who works with adolescents and families. Its mandate, in ACIERA's words, is to “investigate, analyse and communicate” — and Barrón said the work is meant for church ministries “but also to speak to the whole community, to society, to culture and to our country.”

The practical version: a statement can be issued once. An observatory has a mandate and a team, so it has to keep producing. That is the difference between a position and a practice — and it is also the test. The observatory has published nothing under its own name yet, and the launch reports give no budget and do not say whether these roles are paid or dedicated. What matters now is whether it keeps working past the news cycle.

The US bishops: rules for the money, and a task force

The next day, at their fall meeting in Baltimore, the US Conference of Catholic Bishops voted overwhelmingly to add new prohibitions to their investment guidelines, including a section on the types of AI companies deemed ethical enough for investment. The guidelines now prohibit the conference from investing in companies that create autonomous weapons systems or automated targeting systems for the military, security or immigration enforcement without meaningful human oversight; that design AI to be addictive or manipulative, or to harm the mental health of young or vulnerable people; that use AI for genetic manipulation or reproductive technology prohibited by Catholic doctrine; or that create surveillance technology that fails “to prioritize human life and dignity, reinforce biases and prejudices, or violate a person's reasonable expectation of privacy.”

The guidelines go beyond divestment. Through shareholder advocacy, the bishops say they hope to influence AI companies to prevent deepfakes and disinformation, to block pornography, to deter the use of AI in decisions about rationing healthcare or insurance claims, and to promote privacy, workers' rights and environmental protection.

Two honest qualifications belong with that. First, only the bishops' conference itself is bound by the guidelines; many dioceses and other Catholic organisations follow them, but they are not bound. Second, the guidelines are a beginning, not a finished system: the bishops consulted experts in theology and investment, and say the rules “may become more robust” as their understanding develops.

Alongside the vote, the bishops' task force on AI and human dignity, chaired by Bishop Oscar Cantú of San Jose, set out its direction. Cantú cited Pope Leo XIV's teaching document on AI, Magnifica Humanitas, and told the bishops “not to fear the future, but to help shape it, with humanity at the center.” The task force's stated priority is the “democratization of the conversation in the public square” — including listening sessions with Catholics in the pews, labour leaders and educators, and advocacy to Congress, starting with what Cantú called “low-hanging fruit”: child safety, privacy and banking-industry safety. Cantú is also planning webinars in early 2026 to put bishops in front of AI engineers, “to hear directly from the engineers, how the systems are built, how they work.”

That last detail matters. Cantú opened his talk by joking that many bishops had been asking him “whether we're still going to be around in six months.” The anxiety is real, and the task force's answer to it is not a reassurance — it is a room, a conversation, and a list of priorities.

From guidance to government

These two are not the first moves of their kind, and that is the point. Switzerland's churches jointly published ten ethical guidelines, Humanity in the Digital Age, in October — ecumenical persuasion. Austria's bishops adopted the first binding national framework in Salzburg on 12 November: AI in sacraments, preaching, liturgy or pastoral care “may only be supportive,” and replacing human relationships is “inadmissible” — church law in the making, with every diocese to name an AI-responsible person. Argentina adds a standing body with a research mandate. The US bishops add money: what the church will and will not own a share of.

Put together, the pattern is a ladder: guidance, then law, then a team, then the money. Each step is harder to reverse than the one before it, and each one answers a different question — what do we believe, what is forbidden, who keeps watching, and what do we fund.

The practical version: your church does not need to climb the whole ladder. It needs one rung it can actually stand on. For most small churches, that rung is not a committee or a policy document. It is a one-page stance, agreed in one meeting, that says what the church believes about AI and what follows from it.

A one-page stance your church can copy

Pick the lines that are true of your church, delete the rest, and fill in the brackets. Keep it to one page — if it needs a second, it is a policy, not a stance.

Our stance on AI — [Church Name]

What we believe. People are made in the image of God, and no tool may treat them as a means to productivity. Technology must always be subject to ethics; an algorithm cannot replace human discernment, and it cannot replace a person sitting with another person.

What we will use AI for. [For example: first drafts of announcements and the newsletter, captions on the livestream, translation — always reviewed by a named person before anything goes out.]

What we will not use AI for. [For example: sermons, lessons and prayers are written by the person delivering them; we do not enter prayer requests, counselling conversations, giving records or member information into AI tools; we do not use AI that is designed to be addictive, or that transcribes pastoral or children's conversations.]

What we will not buy or support. [For example: we will not pay for a tool whose maker will not say how it uses our data, or one that trains on what we upload without our consent.]

Who decides. [Name / role] is the person to ask about AI in our church, and the person who says no when a new tool does not fit this stance.

Keeping this up to date. We review this stance [every six months / once a year] and whenever we start using a new tool. Last updated: [date].

Notice what the stance borrows from this week's news. The “what we will not buy” line is the US bishops' investment rule, shrunk to the size of a church budget: you may not own shares in an AI company, but you do choose which tools to pay for, and that choice is a rule too. The “who decides” line is Austria's AI-responsible person, shrunk to one name. And the review date is the observatory's lesson in miniature: a stance that is never revisited is a statement, not a practice.

Governing what you publish, too

There is a second half to governing AI that has nothing to do with committees: what goes out under your church's name. A new piece from Amenflow (16 November) argues the AI sermon summary is the riskiest thing a church publishes, because it goes out under the pastor's name and rarely gets a careful read — and offers ten questions to ask any summary tool before trusting it. Condensed, they are:

  1. Can you see the full transcript the summary came from?
  2. Does each quote point to a moment in the sermon, with a timestamp?
  3. Are the “quotes” word for word, or cleaned up?
  4. What happens to a quote the tool cannot find — is it dropped, or printed anyway?
  5. Are Scripture references listed separately from the prose?
  6. Does the tool write out verse text, and in which translation?
  7. Does it separate related passages from the passages actually preached?
  8. Does the summary state the pastor's main point, in the pastor's own framing?
  9. Does it add anything the sermon did not contain — a statistic, a story, an application?
  10. Who presses publish — and has someone who heard the sermon read it first?

The piece compares six tools on how they handle transcripts, quotes and verses — Pulpit AI, Sermon Shots, Pastors.ai, ChatGPT, Otter and Amenflow itself — with features and starting prices as listed on each company's site on 16 November 2026. It is written by a vendor: Amenflow makes one of the six tools, and says so openly. That does not make the checklist wrong — the questions are the right ones whatever tool you use — but it does mean you should run the test yourself: take last Sunday's sermon, check five quotes against the recording, check every Scripture reference against the bulletin, and ask the pastor for the big idea in one sentence. That is a thirty-minute test, and it is governing, too.

The practical version: a stance governs what your church decides. A checklist governs what your church publishes. A small church that has both — one page on the wall, ten questions before anything goes out — has done more governing than most denominations had done a year ago.

An honest limit

Neither of this week's stories is finished, and this piece does not pretend otherwise. ACIERA's observatory has a mandate and a team, but no published output yet and no stated budget — it should be judged in a year by what it has published, not by its launch. The US bishops' investment guidelines bind only the conference, not the dioceses, and the bishops themselves say the rules will need to grow. One forecast in the English write-up of the Argentine launch — that the observatory will publish at least one report, guide or policy proposal on AI's effects on children or adolescents by 30 June 2027 — is the writer's own prediction, given a 45% probability, not a promise ACIERA made; this piece does not repeat it as a plan. The Amenflow checklist is vendor-written and its tool descriptions come from the companies' own pages, not from independent tests. And the Swiss and Austrian pieces this article builds on are linked below, with their own limits stated there.

Sources: Agencia Noticias Argentinas, “ACIERA lanzó el Observatorio de Inteligencia Artificial”, 18 November 2026, and Aire Visión, “ACIERA lanzó el Observatorio de Inteligencia Artificial”, 18 November 2026 — launch (afternoon, 17 November), motto, Hooft's remarks on children and the duty of care, director Ayelén Caffarena and the team, mandate to “investigate, analyse and communicate”; reported in English by Isenfors, “Not on the Altar of Productivity”, 19 November 2026. Religion News Service, “Catholic bishops vote to divest from private prisons, pass guidelines on AI investment”, 18 November 2026 — Baltimore vote, the AI investment prohibitions, shareholder advocacy, the limits (only the conference bound; guidelines may become more robust), Bishop Cantú and the task force, Pope Leo XIV's Magnifica Humanitas. Amenflow, “AI Sermon Summary Tools: A Checklist for Accurate Quotes and Scripture”, 16 November 2026 — the ten questions, the six tools compared, features and prices as listed 16 November 2026, and the disclosure that Amenflow makes one of the tools compared.

Related reading: Five traditions, one voice: the Swiss churches just published the AI guidance the others are still writing — the ten guidelines, Humanity in the Digital Age; Austria just wrote the first binding national church AI rules — AI only supportive, every diocese to name an AI-responsible person; A one-page AI disclosure your church can copy — the congregation-facing statement that follows from a stance like the one above; A church AI policy in a weekend — the staff template; Writing a sermon with AI without losing your voice — the sermon-preparation companion to the summary checklist. This guidance is general and educational, not legal advice; your church's own safeguarding, data-protection, and denominational policies always take precedence.