Safeguarding the human person: AI governance for faith-based investors
- Dave Zellner and Catherine Devitt
- 5 hours ago
- 7 min read
In Part Two of their exploration of AI and faith-consistent investing, FaithInvest's Dave Zellner and Catherine Devitt give their reaction to Pope Leo XIV’s first encyclical, Magnifica Humanitas, and discuss how to turn conviction into governance.
In Part One of this two-part series, we observed that faith traditions are speaking about artificial intelligence with rare clarity, that faith-based investors bring a distinctive voice to that conversation, and that we should already be asking hard questions of the companies whose securities we own.
The more difficult task is to turn conviction into governance, which is the objective of Part Two. We discuss how we think addressing issues surrounding AI belong in an investment policy, what trustees and asset owners should be asking, and how we might structure reporting so that an organisation of any size can confirm what its managers are doing in its name.
Our thoughts on Magnifica Humanitas
Last month, Pope Leo XIV released his first encyclical, Magnifica Humanitas, or 'Magnificent Humanity', which he signed on 15 May 2026 and presented in person at the Vatican ten days later.
He chose 15 May because it marked the 135th anniversary of Rerum Novarum, the encyclical Leo XIII wrote in 1891 to address labour and capital in the upheaval of the Industrial Revolution. Leo XIV draws the parallel deliberately, reading AI as a rupture of comparable scale and bringing the same moral tradition to bear on it.
For investors, one line in the encyclical is notably relevant. Technology, the Pope writes, is never neutral, because it takes on the character of those who 'devise, finance, regulate and use it'. Finance sits squarely on that list, and that placement matters, because through the capital we allocate, we are among the authors of how AI takes shape, not bystanders to it.
In the abstract, technology in and of itself is not a solution to humanity’s problems, just as it is not inherently evil. In practice, however, technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it. – Magnifica Humanitas, 9
The Pope warns against what he calls 'Babel syndrome', by which he means the worship of profit at the expense of the weak, along with forced uniformity and reductionism. And he insists that the real choice is not a simple yes or no to technology but a question of who holds this power and how they use it.
He is candid that AI is not human. It does not experience, suffer, love, or shoulder responsibility, but only imitates the functions of a mind. The task he sets us is to translate principle into practice through responsible planning, an honest assessment of human and social impact, and the deliberate inclusion of the most vulnerable.
This resonates with us as the right way to think about our role. It does not exaggerate the promise of AI, nor does it treat the technology as something to fear in and of itself, and it places responsibility exactly where investors can act – on the governance of power itself.
A shared conviction
This is not a Catholic concern alone. The Vatican first issued its Rome Call for AI Ethics in 2020, and Jewish and Muslim leaders signed it alongside Catholic representatives in 2023. A year later, in Hiroshima, leaders of 11 traditions added their names, bringing Buddhist, Hindu, Zoroastrian and Bahá’í voices to the same six commitments: transparency, inclusion, responsibility, impartiality, reliability, and security.
One Islamic reflection on the Call reads transparency as truthfulness, inclusion as the dignity of every person, and responsibility as moral answerability for power – language that a Reformed, Jewish or Buddhist investor would recognise as close to their own. Roughly 76% of the world claims a religious tradition, according to a 2025 Pew Research Center analysis, and on the question of AI, those traditions are converging on a shared governance vocabulary.
They are converging for good reason. Last month, in our June 2 post, we reported that a research consortium spanning Brigham Young, Baylor, Notre Dame and Yeshiva universities tested every major AI model and found they consistently sidelined religious perspectives, and quietly nudged users toward some faiths and away from others.
The companies in faith portfolios build these very products, and those products can erode the communities we serve, which makes this far more than an abstract worry. It is a concrete question of due diligence.
From conviction to governance
How, then, does conviction become governance rather than sentiment? We think it comes down to three actions for those faith-based asset owners that aspire to align their investments with the moral obligations associated with the use of AI.
The first action...
Name AI in your investment policy. It belongs in the body of the document, not in a footnote to a broader responsible-investment policy – written as commitments you are prepared to hold. In practice, this means doing three things:
Write categorical exclusions into the policy. Name the lines you will not cross by refusing to own companies that, for example:
develop AI for autonomous weapons systems;
deploy AI for oppressive surveillance, or to persecute religious or ethnic groups;
build a business model that commodifies or dehumanises the people it touches.
Define a 'heightened scrutiny' tier for holdings that warrant engagement rather than exclusion. Flag companies for time-bound engagement where you find that, for example:
AI is used in high-stakes settings – healthcare, hiring, credit, criminal justice – without serious governance;
documented algorithmic bias that harms marginalised groups;
opacity that leaves no avenue for appeal or redress;
automation that displaces workers with no credible transition plan.
Set the consequence for each tier in advance. Decide now what triggers an outright exclusion versus what opens a defined engagement window, and how long that clock runs before you escalate or divest – so the policy directs a decision rather than just describing a concern.
The second action...
Apply a simple test that travels across traditions and regulators alike: a responsible system should be fair, transparent, and accountable. Use these three standards as the questions you put to any AI system in your portfolio:
Fairness – does it reinforce bias? Examine whether the data behind a model, and the outputs it produces, entrench existing bias rather than correct for it.
Transparency – can it be challenged from outside? Establish whether anyone beyond the company can understand a consequential decision and contest it.
Accountability – who answers when it causes harm? Identify the named human – and above all the named executive – who will own the harm a system causes, rather than letting responsibility dissolve into 'the algorithm'.
These three standards map almost exactly onto the principles of the Rome Call, which is part of why they travel so easily from one faith to another.
The third action...
Ask the right questions, beginning with your own board and then turning to the asset managers who act on your behalf.
Does our investment policy address AI directly, and have we drawn lines we are genuinely prepared to hold? – a critical question for faith trustees and investment committees
Questions for trustees and investment committees
Does our investment policy address AI directly, and have we drawn lines we are genuinely prepared to hold?
Do we understand where AI already operates across our portfolio, and not only in the obvious technology names?
Have we agreed which uses of AI we will exclude outright and which we will engage?
Would we even learn that one of our holdings had crossed a line we care about, and by what route would that news reach us?
Are we willing to vote our proxies and to file or co-file resolutions on AI governance, rather than simply voicing concern?
How often will you report all this back to us, and in what form will that reporting arrive? – the most important question to ask your asset managers
Questions to ask your asset managers
How do you identify AI-related risks across the whole portfolio, including in companies that no one would label as technology firms?
What would lead you to exclude a company on AI grounds, and have you ever actually done so?
How do you assess a company’s AI governance, from board oversight and bias testing to transparency, redress, and executive accountability?
How do you engage companies on the way they build and deploy AI, and what concrete results can you point to?
How often will you report all this back to us, and in what form will that reporting arrive?
Right-sized, and verifiable
That final question is the one that matters most, because it is what makes everything else verifiable. An organisation does not need a large team or a bespoke framework to govern AI well.
What it needs is more modest and more durable: clear lines written into its policy, a standing place for AI on the agenda of every manager review, and a short annual report from its managers that sets out where AI risk sits in the portfolio, what they excluded or engaged during the year, what changed, and what remains unresolved. A five-person foundation and a multi-billion-dollar retirement fund can both work this way, and reporting is how each of them confirms that its managers are genuinely honouring its convictions.
Yet all of this assumes you can set the terms of a mandate, and many investors, smaller ones especially, cannot. They hold their assets in pooled funds, where the manager answers to many unit-holders at once and cannot bend the portfolio to any single set of convictions.
When you cannot dictate the holdings, your policy is expressed instead in which funds you are willing to own.
That does not put governance out of reach; it relocates it. When you cannot dictate the holdings, your policy is expressed instead in which funds you are willing to own, so the test moves to the front. Before you invest, you ask whether a fund's own AI policy, exclusions and reporting already reflect the lines you have drawn, and you decline the ones that do not.
Thereafter your most consequential act is the decision to stay or to redeem. The questions for managers do not disappear; you simply ask them while choosing a fund rather than after directing one, and you treat a fund's drift away from your convictions as the signal to withdraw.
Pope Leo’s charge to the world was to remain human, and to keep these tools from displacing the human person as we reach for them. For investors, remaining human is not a slogan but a discipline. It means writing down plainly what we will not own, asking our managers to show us that they are listening, and then checking that they are. Our own understanding of AI will keep evolving as the technology evolves, but our obligation to govern it faithfully will not.
Do share your thoughts and concerns about AI – we'd love to hear how your faith organisation is addressing the challenges and opportunities of artificial intelligence.
How faith-consistent investors are navigating artificial intelligence: a special FaithInvest webinar next Thursday
Thursday 23 July
10am New York | 3pm London
We're taking a deep dive into how faith-consistent investors are navigating issues around artificial intelligence next Thursday. And we've got a fantastic line-up of speakers to help us chart a way through the challenges and opportunities of AI. See more below and scroll down to register.
