---
title: "Human First, AI Frontier: Safe and Capable AI After Magnifica Humanitas"
url: https://stacklist.com/card/fb81d8b8-e9f8-4f3b-8232-b4fb8ee90f4c
source_url: "https://miguelguer.substack.com/p/human-first-ai-frontier-safe-and"
stack: https://stacklist.com/c/business/stack/8752ed82-5c8f-4cdb-8f0f-16d30e2b743b
summary: "Miguel Guerrero analyzes Pope Leo XIV's encyclical Magnifica Humanitas as a Christian framework for building AI that is both powerful enough to serve humanity and safe enough not to diminish it. The article argues that capability without safety is Babel and safety without capability is a press release, calling for disciplined deployment and genuine alignment before handing AI decisions we cannot take back."
tags: "ai-safety, encyclical, ai-alignment, human-dignity, catholic-social-teaching, ai-governance, ai-capability"
key_entities: "Miguel Guerrero (person), Pope Leo XIV (person), Pope Leo XIII (person), Magnifica Humanitas (concept), AI alignment (concept), AI safety (concept), Rerum Novarum (concept), The Vatican (organization), Babel vs Nehemiah (concept), catastrophic risk (concept)"
classification: "analysis"
content_hash: "sha256:2d64fd90145ca8622da599b778832b910e5205a9298bd20f580735b07a5ac14a"
acp_version: "0.2"
token_counts_approximate: 2415
visibility: public
agent_accessible: true
status: "final"
---

# Human First, AI Frontier: Safe and Capable AI After Magnifica Humanitas

Human First, AI Frontier: Safe and Capable AI After Magnifica Humanitas Pope Leo XIV’s encyclical is not a rejection of artificial intelligence. It is a Christian call to build AI that is powerful enough to serve humanity and safe enough not to diminish it. Miguel Guerrero May 26, 2026 1 1 Share Article voiceover 0:00 -10:32 Audio playback is not supported on your browser. Please upgrade. I read Magnifica Humanitas twice. Once as a Christian. Once as someone who spends his days where these questions stop being abstract — in evaluations, training runs, deployment rules, and the widening gap between how fast a model ships and how slowly an institution learns to govern it. Both readings landed in the same place. This is the Rerum Novarum of artificial intelligence. Leo XIV signed it on the 135th anniversary of Leo XIII’s 1891 letter on capital and labor — the date is not an accident. The claim isn’t that AI equals industrial capitalism. It’s that the shape of the problem rhymes: a new technological order is reorganizing work, wealth, power, dependency, and truth faster than our institutions can defend the person caught inside it. When that happens, Christians don’t get to retreat into commentary. You have to help rebuild the city. Safe and capable Here’s the frame I’d add to the usual debate. We talk about “ethical AI,” “responsible AI,” “trustworthy AI.” Fine. But the future we actually need is safe and capable. Both words, or neither counts. Capable enough to be worth it — better healthcare, real public services, scientific discovery, education, climate adaptation, an administration that doesn’t lose your file. And safe enough not to corrupt judgment, concentrate unaccountable power, manipulate attention, or quietly turn people into objects to be managed. Capability without safety is Babel. Safety without capability is a press release. This is why I think the encyclical matters more than “the Vatican has concerns about AI.” It refuses the false trade. It never tells us to stop building. It asks what we’re building, for whom, under whose authority, and with what understanding of the human person. And the timing is good. The systems are good enough to matter — coding, math, autonomous operation, scientific work — and not yet reliable enough to trust blindly. Hallucinations, jagged performance, cyber and bio misuse risks, evaluation methods that lag the capabilities they’re meant to measure. That’s the terrain. Not useless. Not safe. In between , which is the hardest place to govern. And the risk doesn’t stay the same size as the system grows. A model that can write code can write exploits; one that can follow a lab protocol can follow a dangerous one. As capability and autonomy scale, so do the failure modes, which is why “safe” has to mean more than catching today’s hallucination. It has to include the harder, still-unsolved problem of ALIGNMENT : keeping increasingly capable systems reliably pointed at what humans actually intend, and under meaningful human control, before we hand them decisions we can’t take back. Catastrophic risk is neither science fiction to be mocked nor doom to be sold. It’s a tail we’re obligated to take seriously precisely because we cannot yet rule it out, and the honest response is neither panic nor acceleration, but disciplined deployment. Babel scales beautifully The encyclical’s central image is Babel against Nehemiah, and once you sit with it you can’t unsee it. Babel is impressive. It coordinates, it scales, it ships, and I’d bet it has excellent metrics. The problem was never capability. The problem was that all that capability pointed at self-assertion instead of communion, a tower to make a name, built without reference to anything above it. That’s the precise warning for AI. A model can be powerful and dehumanizing . A system can be efficient and unjust . An institution can get more automated and less wise at the same time. Babel is the temptation to mistake scale for truth, speed for progress, and control for peace. Nehemiah does something else. He doesn’t perform. He listens, examines the ruins, prays, organizes, shares the work, takes the resistance, and rebuilds with the people, brick by brick. The encyclical sends him straight into the “construction sites of history”: the labs, the companies, the schools, the ministries. That’s the posture I want toward this technology. Not fearful, not merely critical, not waiting for someone else to be brave. The city is broken, so we rebuild it properly, with “prayer” and a threat model, humility and technical competence, the people included rather than optimized from above. The technocratic temptation Leo XIV’s warning about the “technocratic paradigm” isn’t an argument against efficiency but against letting efficiency become the final word on a human life. AI makes technocracy emotionally attractive. Cleaner decisions, faster processing, lower costs, measurable outputs. Some of that is genuinely good: public institutions do need better tools. However, the first question shoult not be “Can we automate this?” but “What human responsibility are we about to change?” Safety isn’t only about preventing the hack or the hallucination. It’s about protecting the conditions under which a person can still think, judge, and refuse. The report I trust most on this is honest about the quiet risks: over-reliance, emotional attachment to systems, and how little we actually know about the long-term effects, especially on children and in education. What happens when kids learn to ask before they learn to think ? When loneliness becomes a business model? When education turns into answer-generation instead of formation? AI should expand human capacity without quietly replacing it, assist judgment without atrophying it. Help people think, not make thinking unnecessary. Regulation is necessary. It is not enough. Europe has done something real with the AI Act. Risk-based, rights-preserving, with full applicability arriving soon. It can make AI safer . It does not make Europe capable . And that’s the trap I worry about most. A continent that elegantly regulates technologies built, trained, and monetized everywhere else is not sovereign, it’s a very well-governed customer. Regulation without capability is dignity on paper. You cannot be the guardian of human-centered AI while depending entirely on someone else to build it, that is coping. So the work cannot stop at rules. It has to build . Compute, talent, evaluation labs, alignment research, multilingual models, public-sector adoption, and institutions credible enough to say no when the evidence is thin and yes when it isn’t. That’s the Nehemiah move: don’t only warn about the ruins, lay the brick, BUILD . Enabling safe AI is itself an act of construction, not an act of restraint. A Europe that only regulates will end up guarding values it no longer has the capacity to defend. Strategies don’t become real by being announced, either. They become real when systems get tested, deployed, audited, broken, fixed, and trusted . That’s hard part, and exactly where Christian realism earns its keep: no naïveté about power, no permission for despair. Good work is possible. It just costs discipline, patience, and institutions. What I’d actually do If I had to compress the encyclical into an operating agenda, it’s shorter than you’d think: Build for dignity, not only for the benchmark. In consequential settings, protect the person, not just the process. Make safety operational. Evaluations, red-teaming, incident reporting, post-deployment monitoring — and the nerve to delay a launch when the evidence is thin. Safety theater is just Babel with a compliance team. Govern the frontier as systemic infrastructure. The most capable models create society-wide risks the moment they ship. Treat them accordingly: shared evaluation, capability thresholds, serious investment in alignment research, and a flat refusal to normalize catastrophic risk as the cost of going fast. Keep judgment human where it counts. Human oversight cannot be a ceremonial checkbox. Define which decisions stay meaningfully ours, and resource the people who hold them. Protect workers before the shock, not after. The market will redistribute the gains and losses. It will not redistribute justice. Labor-impact assessments and reskilling are part of the safety case, not a footnote to it. Treat truth as infrastructure. The danger is not only misinformation, but the slow erotion of a shared reality. Truth stops being a given and becomes infrastructure we have to build. Keep lethal force under human conscience. No machine makes a kill moral. The encyclical’s call to “disarm” AI is the least negotiable line in it. In that order, on purpose The strength of Magnifica Humanitas is that it won’t let us choose between innovation and humanity. It doesn’t ask for nostalgia. It asks for responsibility. It doesn’t tell Christians to leave the technological world. It tells us to walk onto the site. The frontier will be one or the other. Either we build systems that make power faster, more concentrated, and less accountable, or systems that help people flourish, institutions govern, workers adapt, children learn, and truth survive the era of synthetic everything. Nothing about that direction is written yet. It depends on the infrastructure we finance, the systems we evaluate, the launches we have the courage to hold back while empowering the “right” ones, and the institutions we bother to strengthen. This encyclical is an assignment, an assignment to build. But don’t build Babel. Human first. AI frontier. In that order, on purpose. 1 1 Share