Laudate Deum and Laudato Si’ established the pattern Leo XIV follows: technology is not neutral, and enthusiasm for it is not the same as wisdom about it. Magnifica Humanitas sharpens this considerably for AI specifically. The encyclical’s anthropological core is the key. Leo XIV insists that what makes human dignity irreducible is precisely the capacity for relationship, formation, and moral reasoning under uncertainty — the very things AI can simulate but not instantiate. From that foundation, the encyclical doesn’t condemn AI use; it interrogates what kind of use serves the integral development of the human person and what kind quietly hollows it out. The permission structure the encyclical offers is something like: Permitted and even encouraged: AI that extends human reach without substituting for human judgment, relationship, or conscience — diagnostic assistance, research, accessibility tools, translation, administrative burden reduction. Use that frees the human for more human things. Cautioned against: AI that intermediates where personal presence is the point — pastoral care, moral formation, intimate accompaniment. The risk Leo XIV names isn’t malice; it’s the atrophy of capacities we stop exercising because AI does them faster. Structurally condemned: Concentration of AI capability in actors answerable to no community — precisely the subsidiarity violation your bebeholden.org framework addresses. The encyclical reads the governance question as a social justice question, not merely a regulatory one. So the answer to your question is: yes, but the permission is conditional and discerning, not blanket. The faithful are not merely allowed to use AI — they may have an obligation to use it well, as a form of stewardship of the common good. The criterion is always formative: does this use build up human capacity, relationship, and solidarity, or does it erode them in exchange for convenience? That’s the question your elder caregiver clinic is implicitly designed to answer correctly — facilitated, relational, reconnaissance-first rather than tool-deployment-first.