Thinking in Public

 On Magnifica Humanitas: 

What the Encyclical Gets Right, and What It Misses

The moment is significant. Pope Leo XIV's first encyclical, Magnifica Humanitas, intervenes at an inflection point where AI systems are being integrated into consequential decision-making processes, in labour markets, public administration, military systems and information ecosystems, before adequate governance frameworks exist. Its grounding in the tradition of Rerum Novarum is strategically sound: that encyclical emerged when industrial capitalism had already restructured society before labour rights existed. The parallel is more than rhetorical: just as in the Industrial Revolution, technology has already profoundly transformed the conditions of life and work before rights or institutions capable of protecting the most vulnerable people existed. History did not wait for governance, and AI is not waiting either.

However, the encyclical opens with a problematic assertion: that humanity today faces a decisive civilisational choice around AI. That is contestable. Climate change, biodiversity loss, nuclear risk and extreme poverty pose more immediate, more concrete threats to more people. Elevating AI to the status of the central challenge of our era risks inflating its current significance and, more significantly, diverting attention and political capital from challenges where causal chains are better understood and interventions more clearly defined. There is also a structural problem: this narrative of AI inevitability serves the interests of the AI industry, which has every interest in presenting itself as historically unprecedented and civilisationally decisive. The Vatican, even without that intention, ends up reproducing that narrative without questioning it.

The encyclical also does not pose what we might call the question zero: whether certain AI systems should be developed at all, by whom, at what pace and with what democratic mandate. By jumping directly to the question of how to orient development, the document treats the current technological trajectory as a given rather than as a political choice. This absence is not neutral. An encyclical grounded in human dignity and the social doctrine of the Church would have every theological and philosophical reason to affirm that "no" is a legitimate answer, that moratoriums and prohibitions in specific domains are compatible with a Christian vision of progress, and that technological development without democratic mandate is not inevitable but is a choice that can and must be contested. By not doing so, the encyclical misses the opportunity to offer a truly distinctive contribution to the global debate, confining itself to a framework that, in its deep structure, does not differ substantially from the dominant consensus in the industry.

There are nonetheless three things the document gets meaningfully right.

The first is the identification of AI governance as a private power problem. The encyclical is unusually direct in noting that the main drivers of AI development are transnational private actors whose resources exceed those of many states. This is analytically correct and frequently underestimated in political discourse, which tends to focus on regulation as if governments were in a symmetrical position relative to industry. They are not. Most governments lack the technical capacity, the resources and, above all, the political will to challenge the power of the major technology companies. At the same time, multilateral agencies, including the United Nations, find themselves in a state of institutional fragility that prevents them from exercising any effective authority over transnational private actors. The result is a governance vacuum that does not arise merely from regulatory gaps but from a fundamental power asymmetry: governments know they are economically dependent on these companies, fear being left behind in the technological race, and rarely have the political courage to impose concrete limits. Regulation, under these conditions, risks becoming an exercise in legitimation, giving the appearance of control without actually exercising it. The encyclical does well to name this problem. What is missing, inside and outside the Church, is the courage to draw the practical consequences: governments urgently need to shake off their passivity and assume the political responsibility that falls to them.

The second is the treatment of autonomous weapons. The encyclical addresses the military use of AI directly, which is significant in a social encyclical. The normalisation of lethal autonomy, which removes human judgement from decisions about killing, is one of the most urgent and least regulated governance challenges of our time. Framing it within a perspective of human dignity, and not only within humanitarian law, broadens the normative coalition capable of addressing the issue. However, a caveat is needed here that is symmetrical to the one that applies to the rest of the document. Treating military AI as a special category of threat, distinct from other lethal technologies, risks reproducing the same anthropomorphisation error: the idea that it is AI, as such, that constitutes the danger, rather than human choices about its development, its deployment and the absence of democratic control. A precision missile guided by algorithms is not categorically different from other weapons of destruction that already exist and that remain equally poorly regulated. The problem is not the technology; it is the absence of effective human control and political accountability, regardless of the weapons system in question. The encyclical would do better to insist on that point consistently, rather than singling out AI as if it represented an ontological rupture in the military domain that other technologies do not represent.

The third is the attention given to truth as a common good and the call for an ecology of communication. Democratic governance of AI requires an informed public and functioning epistemic institutions. The encyclical correctly identifies the degradation of the information environment as structurally linked to the governance challenge, and not separate from it. However, it stops at diagnosis. The document calls for an ecology of communication and an educational alliance for the digital age, but defines neither of these concepts with any degree of concreteness. There is no guidance on how to regulate algorithmic amplification, how to fund independent public media, how to govern synthetic content, or how to build institutional capacity for digital literacy at the necessary scale. For a document of this ambition, the gap between diagnosis and prescription is notable, and represents a missed opportunity to translate sound principles into useful policy guidance.

Beyond these three points, the encyclical's main contribution is its normative grounding, accessible to audiences broader than specialists. AI governance debates tend to be highly technical or strictly legal. The encyclical offers a vocabulary with genuine cross-cultural resonance, dignity, common good, subsidiarity, solidarity, and has the potential to mobilise communities that conventional regulatory processes do not reach, particularly in the Global South. That is not a small thing.

But two substantial reservations remain. The first, already noted, concerns inevitability. By adopting the Babel/Jerusalem distinction without reservation, the document structurally incorporates the assumption that development proceeds regardless, with governance merely tasked with directing it. That is not a neutral position. It is a political choice that the encyclical should have recognised explicitly as such, rather than presenting it as the natural horizon within which Christian principles apply.

The second concerns anthropomorphisation. The encyclical sometimes weakens its own strongest arguments by describing AI as capable of simulating empathy, wisdom or human relationships. AI systems have no agency, intentionality or moral standing. The real danger is not that machines resemble persons, but that we design institutions and delegation structures as if they did. This categorisation error displaces responsibility from where it actually resides, in creators, operators and regulators, towards a diffuse anxiety about the technology itself.

A concrete example of this framing weakness is found in the section entitled "Remaining human." The title itself is problematic: it presupposes that humanity is at risk of being supplanted or diluted by AI, which simultaneously anthropomorphises the technology and concedes the argument before developing it. For the Catholic Church, whose understanding of the human person encompasses the soul, relationality, embodiment, moral agency and transcendence, the idea that a statistical pattern-recognition system could threaten that reality is as theologically incoherent as it is technically incorrect. The correct framing is not defensive, as if humanity needed to defend its territory against AI, but assertive: AI operates in an entirely different ontological category, and conflating the two is an error, not an inevitable tension to be managed. The risk is not ontological but political and institutional, namely that we choose, through poor governance and misplaced deference, to organise our institutions as if AI were an agent comparable to the human person. The encyclical would have been considerably stronger had it framed this section as a confident assertion of human dignity, rather than a defensive plea for its preservation. That framing would also have made the governance argument sharper: if AI is simply a tool, the question is never about remaining human, but always about who controls the tool, for what purpose and with what accountability.

The circumstances of the encyclical's launch also deserve scrutiny. The presence of Anthropic's co-founder on the presentation panel is symbolically significant and practically ambiguous. On the positive side, it signals that the Church is engaging directly with the builders of these systems, and serious technical engagement is better than none. But Anthropic is a commercial AI developer with its own interests in how AI is regulated. Its association with a normative document of broad reach can be read as legitimation, as if the Vatican were endorsing a specific company's framing of responsible AI. This risk is not hypothetical: industry actors have systematically sought to shape governance narratives in ways that favour incumbents and raise barriers to competitors or to more stringent regulation.

More broadly, the composition of the presentation panel was telling: theology and industry, with no representation from civil society, no critical voices who have worked for decades on responsible AI governance, no representatives of communities that directly suffer the impacts of AI systems, and very limited presence from the Global South. This choice is not neutral. It suggests that the Church's dialogue on AI is being built primarily with those who develop the technology, mediated by theology, and not with those who are governed by it, harmed by it, or have dedicated their careers to holding it accountable. For an encyclical that places the dignity of the most vulnerable at the centre of its argument, this is a notable inconsistency.

The encyclical is stronger as a diagnostic document than as a prescriptive one, which is appropriate to its genre. Its real value will depend on the institutional follow-through the Church generates, namely whether episcopal conferences, Catholic universities and development organisations translate its principles into concrete governance advocacy. Rerum Novarum shaped labour law not through the document itself, but through the organisations and movements it inspired. The same test applies here.

For the encyclical to have maximum impact, the AI governance community should engage with it critically and constructively, particularly on two points. First, the assumption of the inevitability of AI development must be contested: restraint and prohibition are politically legitimate responses in specific domains, and a document grounded in human dignity should say so explicitly. Second, anthropomorphising language must be corrected, not for merely philosophical reasons, but because it displaces responsibility from where it actually resides. The Nehemiah metaphor points in the right direction: human beings deciding, building and assuming responsibility. That logic should govern the framing throughout the document and, where it does not, the encyclical's secular interlocutors should say so clearly and constructively.