Pope Leo XIV’s first encyclical, Magnifica Humanitas, is not a source of international law. It creates no treaty obligations, modifies no customary rules, and settles no legal status for artificial intelligence. However, its relevance lies not only in the fact that it exposes a difficulty that international AI governance has not fully confronted, but also in the fact that it reframes the legal question. It asks whether AI governance is becoming a law of managed technological expansion rather than a law capable of limiting, distributing and directing technological power.

The encyclical belongs to the tradition of Catholic Social Doctrine. In that sense, it is an exercise of papal magisterium: authoritative theological and moral teaching for Catholics, but not a legal instrument for international lawyers. Its argument is built around the res novae of the present, especially digitalisation, artificial intelligence and robotics. It does not treat technology as evil in itself. The point is sharper: technology is never neutral, because it takes on the characteristics of those who design, finance, regulate and use it.

This is why Magnifica Humanitas can matter outside theology. Its vocabulary of dignity, the common good, the universal destination of goods, subsidiarity, solidarity, social justice and peace does not replace legal analysis. It gives international lawyers a diagnostic vocabulary for asking whether law is merely supervising technological systems after their social direction has already been set.

The emerging international and regional AI legal architecture is substantial. The United Nations General Assembly has adopted A/RES/78/265 on safe, secure and trustworthy AI systems for sustainable development. UNESCO has adopted its Recommendation on the Ethics of Artificial Intelligence. The Council of Europe has opened the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law for signature. The European Union has adopted the AI Act. The Global Digital Compact has also moved the debate toward capacity, inclusion and institutional design.

Even when legal and policy instruments invoke dignity, fairness, sustainability and access, their dominant form remains procedural: assessment, oversight and compliance. A compliant AI system may still redistribute authority in opaque ways. The issue is therefore not only whether AI systems can be safe, explainable or supervised. It is whether the global political economy of AI is reorganising power in ways that international law is poorly equipped to control. Magnifica Humanitas matters because it shifts the question from technical governance to domination: who holds technological power, who benefits, who becomes dependent, and whether law merely manages AI’s expansion rather than directs it toward the common good.

From Risk to Domination

Most contemporary AI regulation starts from a plausible premise: AI systems create risks, and those risks must be identified, classified, mitigated and supervised. This is the logic of impact assessment, transparency duties, human oversight and institutional compliance. International law cannot operate without procedures. The difficulty is that the procedure becomes fragile when it is asked to do the work of political economy.

A system may be transparent, auditable, and productive; and still create injustice, dependency, and structural exclusion. Here, the encyclical adds something to the legal conversation. It treats AI not as another technology requiring safeguards, but as part of a broader technocratic paradigm in which private actors, often transnational, hold resources, data and decision-making capacities that may exceed those of many States. In other words, the problem appears when the procedure becomes a substitute for power.

The point is not only ethical but also institutional. States remain the primary bearers of international obligations, while the infrastructures through which AI affects social life are often owned, designed and governed by private corporations. International law does not ignore private power in this context: it addresses corporate conduct through State duties to protect, due diligence, sectoral regulation and standards such as the UN Guiding Principles on Business and Human Rights. Nevertheless, this architecture still presupposes a regulatory capacity that the political economy of AI may itself erode. AI thus intensifies a familiar problem: law addresses States, while practical control is dispersed across corporate networks, cloud infrastructures and privately held expertise.

The question, then, is whether existing safeguards can address concentrated technological authority. The danger is not simply malfunction, but the normalisation of privately mediated governance over work, education, migration, public services, policing, information and political visibility.

The Digital Common Good

The encyclical uses the language of the common good within Catholic Social Doctrine, but the legal point need not depend on confessional premises. In that tradition, the common good concerns the social conditions that allow persons and communities to flourish. Translated into international legal terms, it requires attention to who controls data, who has access to computational capacity, whose languages and histories are represented in datasets, and whose vulnerabilities are transformed into markets.

This matters for the global dimension of AI. International AI governance often presents itself as universal. Its material foundations are not. As Kate Crawford has shown, AI is not immaterial computation but an extractive infrastructure built from minerals, labour, energy, data and logistical systems, none of which is evenly distributed. The Global South may be positioned in many contexts as a source of raw materials, cheap labour, data extraction, and regulatory experimentation while remaining marginal in the design of global standards.

On this view, the digital common good should not be treated as a vague ethical supplement to human rights. In legal terms, it can serve as an interpretive lens that connects existing commitments, such as the right to enjoy the benefits of scientific progress and its applications. In that sense, capacity-building, technology transfer, equitable digital infrastructure and participation in standard-setting are not secondary development concerns, but central to whether AI governance can claim global legitimacy.

The emphasis of the encyclical on subsidiarity is also legally useful. In the digital context, it can be read as a critique of opaque centralisation by major platforms and technology companies. Decisions about access, visibility, content moderation, credit, employment, public benefits, surveillance and mobility should not be dictated by a small number of actors without meaningful participation, contestation and remedy. Glukhin v Russia, on facial recognition and protest, and Schrems II, on transnational data transfers and surveillance safeguards, show that digital infrastructures are sites where privacy, expression, democratic participation and transnational power are legally contested.

This lens also suggests how existing AI norms might develop. Impact assessments should not be limited to individual risk, bias or explainability. They should also ask whether AI systems create structural dependencies, concentrate essential infrastructures, or shift public authority to private systems. Transparency duties should, where appropriate, extend to datasets, computational infrastructures, and supply chains. Capacity-building and technology transfer should be treated not as peripheral development language, but as conditions for equal participation in AI governance. Participation should also extend beyond consultation after design: affected communities, including those in the Global South, should have meaningful opportunities to shape standards before they harden into global templates.

AI, War and the Failure of Legal Minimalism

The clearest point of contact between Magnifica Humanitas and classical international law is war. The encyclical’s concern with weapons and artificial intelligence enters an existing debate on autonomous weapons, human control and accountability. EJIL:Talk! has already hosted important discussions on human control over autonomous weapon systems and on the environmental costs of AI-enabled weapons systems. Those debates correctly insist that human control must be legally meaningful and that AI-enabled warfare has material costs. The encyclical, however, presses a prior question. From Augustine and Aquinas to Vitoria and Suárez, one strand of the just war tradition sought to restrain political violence by subjecting war to criteria of authority, cause, intention, necessity, proportionality and peace. Yet that tradition is also ambivalent. Once war is translated into legal and moral criteria, it can become administrable. The question can shift from whether war should be resisted to whether it can be properly managed.

A significant part of the legal discussion still centres on whether autonomous or AI-enabled weapons comply with existing international humanitarian law. For States parties to Additional Protocol I, Article 36 requires a legal review of new weapons, means or methods of warfare. The UN General Assembly’s recent work on lethal autonomous weapons systems, and on artificial intelligence in the military domain and its implications for international peace and security, shows that the issue is no longer only a technical debate inside IHL.

This is the force of the language of disarmament in the encyclical. It is not limited to the prohibition of a specific category of weapons. It points to a broader problem: the gradual adaptation of law to systems that make violence appear more efficient, precise and acceptable. International humanitarian law is necessary because war exists. International law cannot be reduced to the administration of warfare. The prohibition of the use of force, the duty to settle disputes peacefully and the institutional project of multilateralism all presuppose that peace is not decorative, but a legal and political objective.

AI-enabled violence can be made compliant with the law. Magnifica Humanitas asks whether the technological conditions of violence are making war easier to choose. That question matters for international lawyers.

Conclusion

The encyclical does not provide a legal blueprint for AI governance. Nor should it be treated as if it did. Its contribution is different. It reveals the limits of an international legal imagination that often responds to technological change through proceduralisation. These legal tools are needed, although they cannot answer the prior question of power.

International law contains commitments to dignity, justice, participation, peace and the common good. The problem will be whether those commitments can discipline a technological order increasingly shaped by private infrastructures, geopolitical competition and the promise of efficiency.

The challenge posed by Magnifica Humanitas is therefore not only theological. It is legal and political: whether international law will govern AI as a technology to be optimised or as a field of power to be limited, distributed and directed toward human dignity.