When the Center Speaks for the Periphery: AI Governance and the Limits of Moral Declaration

Contents

On May 15, 2026, one of the oldest continuously operating governance institutions on Earth entered the AI debate. Not with a statement. Not with a press release. With an encyclical.

I. The Document as Institutional Signal

The release of Magnifica Humanitas is not primarily significant for what it says. It is significant for what its existence reveals about the current moment in AI governance.

An encyclical is not commentary. It is the highest form of papal teaching — a document in the tradition of Rerum Novarum (1891), which reshaped labor law and social policy across the industrialized world for decades. When an institution that thinks in centuries, not quarters, chooses this format to address artificial intelligence, the choice of format is itself a diagnostic fact.

The timing is equally revealing. The document did not appear in 2020, when the architecture of AI dominance was still forming and an intervention would have seemed premature. It did not appear in 2030, when infrastructure decisions will likely have hardened into irreversibility. It appeared in May 2026 — at the precise moment when compute concentration, data asymmetries, and model dominance are visible and named, but institutional responses remain contested.

This is what a closing correction window looks like from the outside. A major institution, possessing no enforcement capacity over the technology sector, nonetheless reads the trajectory and decides that the moment for a structural declaration has arrived. The document does not try to regulate AI. It tries to politicize the question of who has the right to regulate it — before that question is answered by default.

That is the first thing to understand about Magnifica Humanitas: it is not a moral appeal directed at individuals. It is an institutional intervention in a contest over normative authority. And it arrives at the last moment when such an intervention can still shift the terms of debate rather than merely document a settled outcome.

Yet there is another question the document’s appearance raises — one the document itself cannot answer: who gets to describe the consequences of AI for those outside the institutions shaping the debate?

II. Three Institutional Signals from the Text

Read structurally, the encyclical contains three signals that matter beyond its theological frame.

The opacity signal (§98). The document states that current AI systems are “more ‘cultivated’ than ‘built’” — that developers do not directly design every detail but create conditions within which the intelligence “grows,” leaving fundamental aspects of its computational processes unknown even to those who build it (§98). This is not a philosophical observation. It is a precise description of the governance problem: you cannot regulate what you cannot audit, and you cannot audit what no one fully understands, including the regulated party. Every governance framework built on the assumption of system transparency — every audit regime, every explainability requirement, every liability structure — rests on a foundation that §98 quietly removes.

The infrastructure signal (§107). Those who control AI, the document states, will impose their own moral vision as the “invisible infrastructure” of these systems. A more moral AI is not sufficient if that morality is determined by a few. This paragraph is not a call for better values among developers. It is a structural diagnosis: the alignment debate, as currently conducted, is a debate among insiders about which insiders should set the terms. The jurisdictions that will live with the results of that debate are largely absent from it. §107 names the mechanism by which this exclusion operates — not through deliberate bad faith, but through the architectural logic of who controls what gets trained on what, and whose edge cases get optimized away.

The extraction signal (§178). The encyclical describes the collection of health data, epidemiological profiles, and genetic maps from structurally fragile regions as a new form of “extractive thinking” — a digital iteration of the colonial resource logic. Those who control this data possess structural leverage over the future: they can shape needs, markets, and allocation before others have defined the questions. This is not a metaphor. It is a description of how predictive model advantage translates into economic and political power. The data that trains the models determining who gets credit, insurance, employment, and healthcare is disproportionately drawn from populations with the least capacity to contest how it is used.

These three signals — opacity, infrastructure capture, and extraction — form a coherent institutional diagnosis. The document reads as if someone had mapped the architecture of AI power and then translated that map into the language of social doctrine.

III. The Contradiction the Document Cannot Resolve

Here the analysis must turn to a structural fact that the encyclical itself cannot address.

The document is built around the principle of subsidiarity: the idea that decisions should be made at the level closest to those who bear their consequences, that centralized authority should not supplant local capacity, that communities must retain the power to ask questions and make corrections. This principle appears in §§68-72 with unusual specificity for a document of this kind — applied directly to digital infrastructure, data governance, and algorithmic decision-making.

And yet the document is itself the product of one of the most centralized normative institutions in the world, issued without any structural mechanism for feedback, correction, or revision by the jurisdictions it describes in §§153 and 178. It speaks about the periphery. It does not speak from it.

This is not a failure of intent. It is institutional physics. Universal moral authority and structural subsidiarity cannot both be maximized simultaneously. The broader the claim to universal authority, the weaker the capacity for local correction becomes. An institution that claims universal normative reach — whether the Vatican, the European Union, the World Bank, or a frontier AI laboratory — inevitably produces knowledge from the center, calibrated by the center’s categories, and validated by the center’s standards of evidence. The populations described in §153 as trapped in hybrid economies, or in §178 as sources of extractable data, have no operational channel through which to contest the document’s framing of their situation, let alone to revise it.

The parallel with the problem the document diagnoses is exact. The encyclical warns that those who control AI will impose their moral vision as invisible infrastructure. The encyclical is itself an act of normative infrastructure — produced by an institution that controls the interpretation of social doctrine, without a formal mechanism through which the periphery shapes that interpretation in return.

This is not a reason to dismiss the document. It is a reason to read it with structural honesty. The encyclical identifies the right problem. But every institution capable of speaking at its scale faces the same structural condition: speaking for others is not the same as giving others a voice. The gap between those two things is where the real work of AI governance still needs to happen.

My position in this analysis is not that of a commentator reviewing a Vatican document from a distance. It is the position of someone who operates in a jurisdiction that §153 describes — where AI is deployed without local institutional capacity to audit it, where the regulator lacks technical expertise to challenge vendor claims, where the court has no precedent and the civil society has no resource. I am not writing about the periphery. I am writing from it. And from that position, the structural contradiction in Magnifica Humanitas is not an abstract observation. It is a description of the gap between what the document recommends and what the institutional reality of most of the world makes possible.

IV. The Question That Remains Open

The encyclical warns against the concentration of moral authority in AI systems. It does not address what happens in jurisdictions where there is no institutional capacity left to challenge that authority — where the regulator cannot audit the system, the court cannot establish precedent, and the civil society cannot sustain an independent challenge.

In those jurisdictions — and they constitute the majority of the world’s population — the question is not how to ensure that AI is governed according to better values. The question is who gets to describe what is happening to the people living under those systems, and whether that description reaches anyone with the power to alter the outcome.

This is the question Magnifica Humanitas raises but cannot answer, because answering it would require the document to proceed from the periphery inward rather than from the center outward. It would require not a declaration about what governance should look like, but an accounting of what governance actually costs in jurisdictions where the institutional substrate for challenge does not exist.

That accounting has not yet been written. It is the work that remains — and it can only be done from positions where the gap between declared governance and operational reality is not theorized from the center, but experienced from within.

Sources & Notes

[1] Leo XIV. Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence. Vatican, May 15, 2026. vatican.va

[2] Leo XIII. Rerum Novarum. Vatican, 1891. vatican.va

[3] UNESCO. Recommendation on the Ethics of Artificial Intelligence. November 2021. unesdoc.unesco.org

[4] Khodjaev, Oybek. Beyond Control: Theory of Limits of AI Governance. Synthesis Memo. Zenodo, May 2026. doi.org/10.5281/zenodo.20120514

[5] Khodjaev, Oybek. The Illusion of Control. Beyond Control series, Essay 1. February 2026. okhodjaev.com

[6] Khodjaev, Oybek. The Agency Transfer. Beyond Control series, Essay 8. April 2026. okhodjaev.com

[7] Khodjaev, Oybek. The Colonial Pattern. Beyond Control series, Essay 5. March 2026. okhodjaev.com

[8] Khodjaev, Oybek. The Correction Window. Beyond Control series, Essay 7. March 2026. okhodjaev.com

Oybek Khodjaev — over 35 years of experience in banking, finance, public administration, and business in Uzbekistan and the CIS. Author of the essay series “Beyond Control: Theory of Limits of AI Governance.” okhodjaev.com

The author advises public institutions and financial organisations on AI governance, verification frameworks, and institutional readiness.

Published