教宗 Leo XIV 的首部通谕《Magnifica Humanitas》于 2026年 5月 25日发布，并把 AI 描述为现代日常生活的「隐形基础设施」。文本刻意回望 Leo XIII 在 1891年 5月 15日发表、于其 135周年被重新援引的《Rerum Novarum》，把当年的工厂、劳工与工业资本主义，对照今日的平台、演算法、资料与自动化系统，强调科技并非天然邪恶，但其规模与深度已使「人类从未如此掌握自身的力量」成为现实。
通谕的核心是「解除武装的科技」：不是阻止 AI 发展，而是避免少数全球行为者在最高性能演算法与最大资料中心的竞赛中，集中数位基础设施、资料与运算能力，进而影响资讯、经济、民主与权力。它同时指出，平台与演算法往往依最大化注意力与互动来筛选资讯，使可见内容未必最真实，而是最能引发反应；因此公共判断不能只交给以市场或权力逻辑运作的数位基础设施。
在工作面向上，Leo XIV 警告技术失业可能引发「社会灾难」，因为为了降成本与增利润而推动的自动化，会让劳工被压缩为可重复、可控制、可替代的功能，并伴随监控、任务碎片化与自主性丧失。在战争面向上，通谕拒绝把致命或不可逆决策交给人工系统，认为当代冲突已被自动化系统深度渗透；因此，真正的问题不是 AI 本身，而是谁控制它、为何而用，以及它正在建构何种人类与社会秩序。
Pope Leo XIV’s first encyclical, Magnifica Humanitas, was published on May 25, 2026, and frames AI as part of the “invisible infrastructure” of daily life. It deliberately recalls Leo XIII’s Rerum Novarum, issued on May 15, 1891 and revisited on its 135th anniversary, to contrast 19th-century factories, labor, and industrial capitalism with today’s platforms, algorithms, data, and automation systems. The text argues that technology is not evil in itself, but its scale and depth now make it true that humanity has never had so much power over itself.
Its central idea is “disarming technology”: not stopping AI, but preventing a few global players from concentrating digital infrastructure, data, and computing power through a race for the highest-performing algorithm and the largest data center. That concentration affects information, economics, democracy, and governance. The encyclical also says platforms and algorithms filter reality by maximizing attention and engagement, so what becomes visible is not necessarily what is most true; public judgment should not be left to digital infrastructures driven by market or power logics.
On work, Leo XIV warns of a “social calamity” if technological unemployment is driven by cost-cutting and higher profits, because workers can be reduced to repetitive, tightly controlled, replaceable functions, with surveillance, task fragmentation, and lost autonomy. On war, the encyclical rejects delegating lethal or irreversible decisions to artificial systems, arguing that automated conflict already distances decisionmaking from human responsibility. Its final point is that the real issue is not AI as a technical object, but who controls it, for what interests, and what kind of human and social order it is helping build.