Holy See (Vatican – central governing body of the Roman Catholic Church) - Encyclical Letter Magnifica Humanitas of his Holiness Pope Leo XIV – On safeguarding the human person in the time of AI

In Magnifica Humanitas, Pope Leo XIV issues his first systematic engagement with artificial intelligence as a governance question rather than a series of discrete ethical cases. Published on the 135th anniversary of Rerum Novarum, the encyclical mirrors Leo XIII's response to industrial capitalism: it names the transformation of its era, grounds a response in principles of human dignity, and calls on political and economic actors to use technological power to the common good.

The document spans five chapters and covers topics such as the development of Catholic Social Doctrine, the theological and philosophical foundations of the human person, the specific challenges of AI, the protection of truth and democratic life, labour in the digital transition, and the relationship between technological power and the possibility of peace. Its scope is unusually broad even by encyclical standards because it addresses topics like autonomous weapons, epistemic fragmentation, platform dependency, and the labour market alongside the more familiar questions of data ethics and algorithmic decision-making. It is also very ambitious: it does not offer a technical framework but a moral architecture, a set of orienting principles that it calls on legislators, designers, educators and civil society to operationalise.

The encyclical should not be read as a policy document in the strict sense. Its primary purpose is theological and doctrinal rather than regulatory or programmatic. This matters for the analysis: while the text can offer important normative guidance and shape public debate, it does not set out concrete policy instruments, implementation pathways, or institutional responsibilities in the way a formal policy document would. Any policy relevance therefore needs to be derived from its broader ethical and theological principles, rather than from direct prescriptions.

What: Papal encyclical letter on AI, human dignity and the common good. Policy orienting document

For whom: Catholic faithful, all Christians, all people of goodwill; also addressed to legislators, technologists

URL: https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html 

Relevance for Belgian and Flemish stakeholders:

• The encyclical's treatment of labour displacement is perhaps where the resonance is sharpest. Belgium's system of sectoral collective bargaining, administered through the National Labour Council (NAR/CNT), provides exactly the kind of institutional channel through which the document's argument (that displacement is a distributional choice rather than a technical inevitability, and that workers must have a voice in how productivity gains are allocated) can be translated into practice. The Joint Industrial Committees (paritaire comités) that negotiate sectoral agreements are already engaging with AI and automation questions; the encyclical's insistence on solidarity and the dignity of work (§156) gives those negotiations an explicit normative frame they can draw on, or push back against, depending on the interests at the table.
• Belgium's media landscape adds a second dimension. The encyclical's call for an "ecology of communication" to counter epistemic fragmentation caused by algorithmic curation describes a dynamic that takes on a specific character in a small, multilingual country. Platform recommendation systems, trained predominantly on English and large-language content, tend to pull users toward dominant-language information environments, with consequences for the coherence of distinct Flemish and Francophone public spheres. The fragmentation the encyclical worries about is not only ideological, it is also, in the Belgian case, linguistic and communitarian. This is a research question as much as a policy one, and one that Belgian media scholars and the Flemish and French-language regulators (VRM and CSA respectively) are positioned to take up.
• Finally, Belgium retains a dense Catholic institutional presence whose relevance the encyclical explicitly activates. A network of Catholic schools, hospitals and social welfare organisations across both language communities constitutes a point of potential mobilisation that secular governance actors do not possess. The encyclical does not merely address these institutions as its audience; it assigns them a specific role in the "shared discernment" it calls for. Whether they take up that role in a way that produces concrete governance contributions, rather than symbolic affiliation, remains to be seen.

The release of Magnifica Humanitas arrived at a specific political moment. Across much of the world, formal governance initiatives for AI have stalled, softened or been rolled back. In the European Union, the AI Act’s promise of binding oversight was being simplified through the Digital Omnibus, which postponed key obligations for high-risk systems and recast delay as simplification. In the United States, under the influence of voices in Silicon Valley and Washington who framed safety advocacy as an obstacle to competitiveness, several oversight structures assembled during the previous administration had been dismantled. At the same time, the debate over federal AI regulation shifted toward pre-emption: rather than building robust protections, policymakers increasingly considered whether state-level safeguards should be frozen in favour of a lighter national framework for frontier AI developers.

Into this environment, the encyclical intervenes from an unexpected direction. The papacy is not a technology regulator. It has no enforcement competence over technology companies and no direct role in the legislative processes through which AI governance frameworks are adopted. At the same time, the Holy See is not simply a private moral voice: it has a recognised international legal personality and can enter into international agreements. What it has is reach: roughly 1.4 billion Catholics worldwide, a sophisticated international diplomatic presence, and a tradition of social teaching that has, at its best, provided orienting frameworks for political debate well beyond the Catholic Church. Magnifica Humanitas is an attempt to use that reach in the AI debate, at the moment when the public conversation is increasingly dominated by those who favour acceleration over restraint.

The document is structured as follows: chapter one establishes the dynamic, historically-embedded character of Catholic Social Doctrine. Chapter two lays out its foundational principles (human dignity, the common good, subsidiarity, solidarity, social justice) as the lens through which to read the rest of the document. Chapter three addresses AI directly, by distinguishing between transhumanist and posthumanist ideologies on the one hand and a Christian humanism that refuses to reduce the person to data and performance on the other. Chapter four addresses truth, work and freedom as specific domains threatened by the current trajectory of technological development. Chapter five addresses the culture of power, including a section on autonomous weapons and the failure of multilateral institutions.

Core arguments and their governance implications

The document’s most important contribution to the governance debate is its insistence that technology is never neutral, a claim it makes with unusual precision. Rather than stopping at the familiar observation that AI systems can produce biased or discriminatory outcomes, the encyclical identifies the structural condition that makes this the case: technology takes on the characteristics of those who devise, finance, regulate and us it. The implication is that governance cannot function as an ex-post correction applied to systems already built. It must be embedded in the design process from the outset.

The Babel/Jerusalem frame
 

The theological structure of the document rests partially on two contrasting biblical images: the Tower of Babel and the rebuilding of the walls of Jerusalem under Nehemiah. This is not just decoration. The Babel story is read as a template for technological overconfidence: the construction of a project oriented toward self-glorification, homogenization, and the elimination of human diversity in the name of efficiency. The Nehemiah narrative, by contrast, models a form of collective action: distributed, participatory, oriented toward rebuilding social bonds rather than asserting domination. The encyclical uses these images as normative frameworks. AI development oriented toward the “Babel syndrome” (the idolatry of profit, the reduction of the person to data, the aspiration to resolve the mystery of human subjectivity through technical optimization) is explicitly condemned. Development that instead supports pluralism, enables participation, and measures progress by the dignity of each person is affirmed.

What makes this framing more than rhetorical is its specific application. The encyclical does not treat the choice between Babel and Jerusalem as abstract. It names the structural features of the current moment that make the Babel outcome more likely: the private, transnational character of the main drivers of AI development; the asymmetry in resources and capacity between technology companies and governments; and the tendency for the succession of technological emergencies to prevent sustained reflection on direction and purpose.

The five thematic domains

The encyclical organises its normative interventions around five themes, each with identifiable governance implications:

*Here, however, “human dignity” should be understood in its theological context. The encyclical does not use dignity only in the contemporary policy sense of equal status, autonomy, or protection against discrimination. It also refers to the dignity of the human being as a created being, whose life is oriented toward a moral telos. This gives the term a different normative foundation from many present-day policy debates.

On power concentration

One of the document's most analytically sharp observations concerns the shift in the locus of technological power. In the past, the encyclical notes, it was largely up to states to guide and direct innovation. Today the main drivers of development are private, often transnational actors whose resources and capacity to intervene exceed those of many governments. This is framed not as an inevitable feature of technological development but as a specific political condition — one that makes it "even more challenging to discern, govern and direct such power toward the common good."

The encyclical names this condition with unusual directness: it denounces "new AI monopolies" and the "epistemic, economic, and political asymmetry" they create (§109), and insists in paragraph 110 that it is not enough to regulate AI, it must be "disarmed and made accessible." This is the document's single most striking normative claim, and one that goes well beyond the vocabulary of most governance frameworks. "Disarm" is deployed deliberately and repeatedly across the text. It is applied to words, to weapons and to commercial competition logics. Its application to AI development signals that the encyclical's concern is not only with how AI is used but with the structural conditions under which it is built, financed and controlled. The observation that governance cannot reach what it cannot see ( that the moral imprint of AI is fixed at design stage by actors whose accountability is primarily to capital) carries a direct implication the text does not shy away from: a more moral AI is useless if that morality is decided by only a handful of people (§107).

Transhumanism, posthumanism and the simulated self

Chapter Three includes a sustained engagement with transhumanist and posthumanist ideologies that is unusual in a governance document. The encyclical identifies these currents not merely as philosophical positions but as ideological underlying foundations of specific product visions and investment rationales: the aspiration to enhance or transcend human biology through technology, to treat human limitation as a design flaw rather than a constitutive feature of the person. Against this, the document insists on a Christian humanism that refuses to equate progress with the elimination of vulnerability. Weakness, limitation, and dependence are not errors to be corrected; they are dimensions of the human person that call for care, solidarity and relationship rather than technical resolution.

The encyclical's most precise intervention in this territory concerns the simulation of relationship. Paragraphs 99 and 100 make a distinction that deserves to be taken seriously beyond its theological context: AI systems "do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean." They may imitate language and simulate empathy, but they do not understand what they produce. The risk the encyclical identifies is not, primarily, that people will be deceived into thinking they are speaking with a human, it is subtler: "that he loses the very desire to seek the other authentically" (§100). The concern is not confusion but attrition, meaning the gradual displacement of the impulse toward genuine encounter by the availability of its optimised simulation. This is a claim about what dependency-by-design does to human sociality over time, and it is one that governance frameworks have barely begun to address. 
 

The encyclical was received with notable breadth of engagement across the tech policy community. Commentators across different analytical traditions found points of resonance, though several also identified significant tensions and gaps.

On the governance deficit

The most pointed critical observation concerned the distance between the document's normative ambitions and its institutional mechanisms. The encyclical sets moral parameters with clarity but leaves open the central question of where ultimate responsibility lies: with users, with designers, with states, with supranational bodies? The document's structural indeterminacy is partly principled: the logic of subsidiarity resists top-down specification of institutional arrangements. But it also means that the encyclical can be simultaneously endorsed by people who hold radically different views about whether governance should be designed at platform level, member state level, or through international treaty.

On labour and the distribution question

Chapter Four, the longest section of the document, is also the most concrete. The encyclical’s treatment of work arrives at a moment when the labour displacement question has become impossible to avoid. US companies announced over 300,000 job cuts through April 2026, with AI cited as the primary driver by many employers. The document cuts through the marketing language around AI and productivity with unusual directness: contrary to the promises made on its behalf, AI can “paradoxically deskill workers, subject them to automated surveillance, and relegate them to rigid and repetitive tasks” (§150). Displacement is not treated as an inevitable feature of technological development but as a choice: one that reflects specific decisions about how productivity gains are allocated and whose interests innovation is made to serve. The encyclical responds by afferming state regulation, labour unions, and “social criteria for innovation” (§156) as legitimate instruments, and situates them within the Catholic Social Doctrine tradition of worker solidarity that runs from Rerum Novarum through to the present.

The encyclical does not engage with the specific policy instruments such as wage subsidies, universal basic income, labour transition funds, instruments through which the distributional question might be operationalised. But its normative framing provides a foundation for those debates that is harder to dismiss than purely technocratic instruments: an economy that treats workers as replaceable inputs, substituting labour for capital wherever the calculation is favourable, has already made a moral choice that the document holds to contradict human dignity at its root.

On truth, algorithms and the information environment

The encyclical's engagement with epistemic fragmentation is notably fluent. The observation that algorithmic curation shapes not only what people know but how they know, what they consider plausible, and who they trust, sits within a research tradition in media studies and communication science that has been tracing these dynamics for over a decade. The document calls for an "ecology of communication"  and an "educational alliance for the digital age." These are real institutional proposals, even if their operationalisation is left to others. The call for "media literacy" as a civic competence has been a staple of digital governance discourse for years; the encyclical adds to it the observation that literacy must extend to the recognition of algorithmic mediation itself: that the question is not simply what information people receive but through what structures and with what optimisation targets.

On the spiritual dimension as governance blind spot

Perhaps the most interesting critical observation came from those who argued that even the encyclical itself understates one dimension of the AI problem. Danielle A. Davis Canty for example identified what she called the "sacred data" question: the entry of AI systems into the spaces of prayer, confession, moral discernment and spiritual guidance. These are domains where deeply personal expression becomes computational data, where the relational and communal contexts that historically gave such expression meaning and protection are absent, and where existing privacy law offers no special protection. This critique is relevant for policy beyond its theological dimension: it identifies a category of human vulnerability (intimate expression in contexts of trust) that current data governance frameworks do not adequately address.

On autonomous weapons

The encyclical's treatment of autonomous lethal systems is its most unambiguous policy position. It calls for an international ban on weapons systems capable of making lethal decisions without human moral accountability in the loop. This position aligns with the campaign for a Lethal Autonomous Weapons Treaty at the UN level, which has gained traction in recent years but has been resisted by major military powers. The encyclical does not engage with the technical question of what constitutes "meaningful human control" but it insists on the moral non-delegability of lethal decision-making with a clarity that is unusual in official documents from any institutional source.
 

The encyclical's most important contribution to the governance debate is structural rather than substantive. By framing AI governance as a question of shared discernment, requiring the participation of "scientists and researchers, entrepreneurs and workers, educators and legislators, civil society, popular movements and faith communities", it establishes a multistakeholder model grounded not in procedural efficiency but in the moral claim that decisions affecting the common good require the participation of those affected. This is a strong position: it challenges both the techno-optimist argument that governance is unnecessary and the technocrat argument that governance is a matter for specialists.

The document is weaker on the institutional mechanisms through which this participation might be organised. It invokes subsidiarity and solidarity as governing principles but does not specify the forms of democratic oversight through which they might be realised. This is partly a feature of the genre: encyclicals are not legislative drafts. But it means that the document's normative force depends entirely on whether it can catalyse institutional actors to supply what it omits.

The question of whether it can do so is genuinely open. The Catholic Church's global network constitutes a form of institutional reach that secular governance actors do not possess. If the encyclical succeeds in mobilising even a portion of that network in support of specific governance initiatives, it will have done something that no comparable document has managed in the AI debate: provided a moral architecture with real organisational backing. Whether this happens will depend on translation: the difficult work of connecting the document's theological language to the operational requirements of regulatory design, procurement standards, platform accountability and labour law.

Frederic Heymans