Human First, AI Frontier: Why Magnifica Humanitas Is the Rerum Novarum of Artificial Intelligence

The question is not whether AI will be good or bad. It is what kind of human order we are building with it — and who gets to decide.

AI is no longer a laboratory promise. It organizes markets, mediates education, reconfigures war, concentrates computing power, and changes the conditions under which truth itself is produced and contested. The usual question — will AI be good or bad — is the wrong one. It asks about a tool. We are not facing a tool.

Leo XIV places the encyclical in explicit continuity with Rerum Novarum, Leo XIII's 1891 encyclical on capital and labor, whose 135th anniversary falls this year [1]. That continuity is the argument. Rerum Novarum responded to a world in which industrial capitalism had reorganized work, wealth, and dependency faster than existing institutions could absorb. Magnifica Humanitas addresses a comparable mutation: the shift from industrial capitalism to a capitalism of data, models, platforms, compute, and cognitive infrastructure.

My thesis is simple: this encyclical should be read as the Rerum Novarum of artificial intelligence — not because it repeats the past, but because it identifies the same structural problem. When a new technological order reorganizes economic and social life faster than institutions can defend dignity, politics has to rethink its foundations. Everything else is commentary.

This essay continues a thread I have been pulling on in Europe's Moment to Lead on the political economy of frontier AI, Governance as Advantage on institutional capacity, and Spain Is Not a Translation Problem on building AI that understands the country it serves.

AI Is Not a Tool. It Is Organized Power.

The first strength of Magnifica Humanitas is that it refuses to treat technology as neutral. The text does not demonize AI; it recognizes that technology can heal, connect, educate, and expand human capability. But it insists on what is easy to forget in a launch cycle: technology always takes the shape of the incentives, interests, institutions, and worldviews of those who design, fund, regulate, and deploy it [1].

From inside the institutional machinery, that is obvious daily. AI is not just software. It is data, talent, chips, energy, cloud infrastructure, models, interfaces, benchmarks, procurement contracts, legal obligations, market incentives, and design choices — power, in a single word. The encyclical names this without euphemism: the central drivers of technological development are no longer mainly states but private, often transnational actors with resources and operational capacity that surpass those of many governments, which makes that power harder to see, govern, and direct toward the common good [1].

This is not abstract. Stanford's 2026 AI Index describes a landscape in which industry produced more than 90% of notable frontier models in 2025, organizational AI adoption reached 88%, and the United States hosts 5,427 data centers — more than ten times any other country. U.S. private AI investment reached $285.9 billion in 2025, over 23 times China's $12.4 billion [2].

Source: Stanford HAI, 2026 AI Index Report [2].

Source & caveat: Stanford HAI, 2026 AI Index Report [2]. Private-investment figures likely understate China's total spending because state guidance funds are excluded; the performance gap narrowed from double digits in 2023 to roughly 2.7% by March 2026. Capital asymmetry is real; technological leadership is far more contested than the “U.S. vs. China” capital narrative suggests.

But here is where serious AI policy has to apply its first rule and interrogate the headline number. The same report notes that private-investment figures likely understate China's total spending because of state guidance funds, and that despite the 23-to-1 capital asymmetry, the performance gap between the leading U.S. and Chinese models had narrowed to roughly 2.7% by March 2026 [2]. Concentration is real; infrastructure power is real; but technological leadership is far more contested than the “U.S. versus China” capital narrative suggests. For Europe, that means two things at once: the window is open, and the window is narrow.

Europe cannot afford to be merely the continent that regulates AI systems built, trained, deployed, and monetized elsewhere. Regulation is necessary but not sufficient. The strategic question is whether Europe — and Spain within it — can build its own architecture of trust: auditable systems, multilingual models, sovereign compute, shared data, credible supervision, skilled talent, responsible procurement, and democratic legitimacy. A human-centered AI cannot remain a value statement. It has to become infrastructure.

Babel or Nehemiah

The encyclical's most powerful image is the contrast between Babel and Nehemiah. Babel is the tower of uniformity, control, and self-sufficiency — the fantasy of a single language, a single direction, a single technological order that sacrifices human dignity for scale. Nehemiah is the patient rebuilding of a city: shared responsibility, plural voices, concrete work, repaired relationships, a community capable of inhabiting its future together [1].

I should be honest about what I am doing with that image. I read it as political philosophy; Leo reads it as doctrine, and for him Nehemiah's rebuilding is inseparable from God at its center. That is a real difference, and I am borrowing the metaphor while leaving part of its content behind. But the structure survives translation, because it describes two genuine models of how to build AI.

This is where Magnifica Humanitas moves the debate in the right direction. The question is not only which model wins a benchmark. It is who can audit it, who depends on it, who is excluded by it, who captures the value, who bears the risk, who can contest the decision, and who answers when something goes wrong. That is the governance question — the one that separates serious AI policy from technological theater.

The Technocratic Temptation

Leo XIV returns to a critique that runs through modern Catholic social thought: the danger of the technocratic paradigm. This is not an argument against efficiency, but against allowing efficiency to become the final measure of every human decision. When control, performance, and profit become the dominant language of institutions, people are gradually reduced to components in a system. The problem is not that technology is powerful; it is the belief that what is most powerful is therefore what is best.

That distinction is essential for AI. A public administration should not adopt AI merely because it cuts costs, a company should not automate merely because it improves margins, a school should not introduce generative systems merely because they accelerate output, and a hospital should not deploy an algorithm merely because it optimizes a workflow. The first question is never “what can we automate?” It is “what relationship, right, responsibility, or form of judgment are we altering?” AI does not remove the need for judgment; it makes judgment more important.

Work: Dignity Cannot Be Automated

The most important section of the encyclical is the one on work, and it is where the link to Rerum Novarum stops being a metaphor. Leo XIII wrote in a world where the factory had reorganized the relationship between capital and labor; Leo XIV writes in a world where AI, robotics, and automation are reorganizing not only employment but professional identity, autonomy, learning, creativity, and the social value of human contribution. Work is not merely an economic input — it is one of the places where dignity becomes visible. A society that grows more productive while making work more precarious, opaque, surveilled, or meaningless is not becoming more intelligent. It is only becoming faster.

The public debate swings between two caricatures: infinite productivity on one side, mass unemployment on the other. Reality will be messier. Some jobs will vanish, others will be transformed, new professions will emerge, and many tasks will be recomposed around human-machine systems. But one thing is already clear: without institutions, the transition will be regressive.

Source & caveat: Stanford HAI, 2026 AI Index Report — Economy [3]. The fall in entry-level software employment cannot be pinned cleanly on AI alone — hiring cycles, interest rates, and post-pandemic corrections all matter, and the report notes that senior-developer headcount kept growing. Direction is clear; clean causal stories are not.

The Stanford AI Index reports organizational AI adoption at 88% in 2025, with generative AI used in at least one business function by 70% of organizations, and employment for software developers aged 22 to 25 down nearly 20% since 2024, even as one-third of surveyed organizations expect workforce reductions in the year ahead [3]. The right response is precision, not panic. That fall in entry-level software employment cannot be pinned cleanly on AI alone — hiring cycles, interest rates, and post-pandemic corrections all matter, and the report itself notes that senior-developer headcount kept growing. Anyone selling a clean causal story is selling something. But the direction of policy is clear: major AI deployments should arrive with labor-impact assessments, worker participation, reskilling, sectoral transition agreements, and metrics that measure not only productivity but job quality, autonomy, and dignity. The market will redistribute gains and losses; it will not redistribute justice on its own.

Truth and Interior Freedom

In the age of generative AI, truth is no longer only an epistemological problem. It is democratic infrastructure. When text, image, audio, and video can be generated instantly and cheaply at scale, the problem is not only misinformation but the erosion of public trust. If everything can be fabricated, everything can be denied; if everything can appear authentic, authenticity loses political force; and if public attention is organized by opaque recommender systems, democratic conversation comes to depend on architectures most citizens cannot see and most institutions cannot fully govern [5].

This is one of the most political parts of the encyclical. Freedom does not only mean the absence of coercion; it also means the capacity to pay attention, deliberate, remember, judge, form relationships, and resist manipulation. A society can remain formally free while becoming psychologically captured by systems designed to optimize behavior. That is why AI should not be analyzed only through privacy, cybersecurity, or compliance, but also through interior freedom.

A society can remain formally free while becoming psychologically captured by systems designed to optimize behavior.

What happens when children learn to converse with synthetic systems before they learn to sit with silence, conflict, or frustration? When loneliness becomes a monetizable interface? When education becomes automated answer-production rather than the formation of judgment? The encyclical warns that simulated empathy is especially dangerous where real relationships are already scarce, and that AI can weaken judgment when it rewards excessive delegation [1]. None of this is an argument against AI in education, health, or public services. It is an argument for designing it with limits: AI that expands capacity without replacing relationships, assists judgment without atrophying it, and frees time for what is human rather than colonizing the space where human formation happens.

War: No Machine Makes a Lethal Decision Moral

One of the hardest sections concerns war. The warning is direct: autonomous systems can lower the moral threshold for the use of force by distancing human beings from the visible consequences of violence, and decisions that are lethal and irreversible should not be delegated to artificial systems — there must be traceability, effective human control, accountability, and international rules [1].

This is not science fiction. The 2026 International AI Safety Report, backed by some thirty countries and international bodies, analyzes risks from general-purpose systems — impacts on human autonomy, concentration of power, labor markets, and the rising capabilities of increasingly autonomous systems — and is blunt that risk management remains uneven and immature [6]. The algorithmic transformation of war forces us to recover a basic principle: not everything technically possible should be politically permitted. Automating violence does not remove moral responsibility; it redistributes it, obscures it, and often makes it easier to evade. Europe should lead with an unambiguous position — meaningful human control, a prohibition on full lethal delegation, international auditing, and stronger protection of civilians and critical infrastructure — instead of regulating commercial AI while looking away from the military kind. The most dangerous uses of AI will not always be the most visible.

Europe: Regulation Is Necessary, but Insufficient

Europe has done something important with the AI Act, the first comprehensive legal framework for AI: it entered into force on 1 August 2024 and becomes fully applicable on 2 August 2026, with phased milestones along the way — earlier dates for prohibited practices, AI-literacy obligations, and general-purpose AI rules [4]. But a legal framework is not a technological strategy. Europe cannot become only the continent that classifies risk while others build the models, own the infrastructure, set the interfaces, capture the data, and monetize the dependency. The European opportunity is larger than regulation: an ecosystem that is trustworthy, auditable, multilingual, interoperable, sectorally useful, rights-preserving, and institutionally legitimate.

Source: European Commission, “AI Act — Regulatory Framework” [4]. A legal framework is necessary but not sufficient. The next phase is implementation, supervision, and operational uptake.

Here I owe the reader a disclosure: I work inside Spain's responsible-AI program, so I have every incentive to tell a heroic story about what we are doing. I will try not to. The Spanish AI Strategy 2024–2025 is backed by €1.5 billion, on top of €600 million already mobilized, and is structured around strengthening capabilities, applying AI across the public and private sectors, and fostering transparent, human-centered AI [7]. AESIA, the Spanish supervisory agency, is tasked with supporting ethical, safe, rights-respecting AI and overseeing compliance [8]. And ALIA is an open family of Spanish and co-official-language models meant to reduce dependency and provide public, accessible linguistic infrastructure [7].

Sources: Plan de Recuperación — Estrategia Española de IA 2024 [7]; AESIA [8]. Disclosure: the author works inside Spain's responsible-AI program.

This is the right terrain — but I should concede the skeptic's point before answering it. Europe has been promising “trustworthy AI” for the better part of a decade, and it has more strategies, declarations, and frameworks than it has adoption to show for them. Declarations do not build trust; press releases do not become infrastructure; strategies matter only once they are implemented, evaluated, audited, and actually used. That is the hard part, and it is precisely where Spain should raise its ambition rather than rest on its documents. The goal cannot be merely to “use AI” or to “comply with the AI Act.” It should be to demonstrate that a human-centered AI can be more adoptable, more trustworthy, and more useful precisely because it is built around institutions, rights, languages, workers, public services, and real economic sectors. Europe's advantage will not be the largest model. It must be the most legitimate ecosystem.

A Human First Agenda for the AI Frontier

If we take Magnifica Humanitas seriously, the response cannot stop at interpretation. It has to become an agenda. Seven points. None of them new on their own; together they describe what a Human First frontier looks like in practice.

That is what “Human First, AI Frontier” should mean. Not anti-technology, not slow technology, not symbolic ethics — advanced AI that is also auditable, adoptable, democratic, multilingual, useful, responsible, and governed.

Against Human Abdication

The force of Magnifica Humanitas is that it refuses the false choice between innovation and humanity. It does not ask us to stop AI but to govern it; it does not ask for nostalgia but for institutions; it does not ask for fear but for responsibility. That is why the Babel/Nehemiah contrast holds. Babel is AI that rises without looking down: more scale, more speed, more dependency, more opacity, more power without accountability. Nehemiah is the patient work of rebuilding a city that can actually be inhabited — rights, work, education, truth, peace, language, public institutions, and human bonds.

Europe and Spain have a responsibility here that goes beyond being users or regulators. We have to build a frontier of our own: one where AI is not only powerful, but legitimate. The direction of this technology is not written. It will depend on the infrastructure we finance, the rules we enforce, the models we build, the workers we protect, the children we educate, the data we share, the limits we set, and the institutions we have the courage to create.