Leo XIV warns: letting algorithms decide for us drains human relationships of substance

By Vaticano · September 16, 2026.

During the general audience in St. Peter's Square, Pope Leo XIV warned that artificial intelligence tools can make human relationships "increasingly less personal, more virtual," and urged the faithful not to get used to "delegating to algorithms the decisions that are the exclusive preserve of the human being," according to the papal catechesis. The Pontiff qualified that technology and social media do help shorten the distance between people, but stressed that moral decisions cannot be outsourced to an automatic system: "Ensuring that social relationships have real substance and personal depth is, in this sense, the contribution the Christian community seeks to offer," he said.

It is not the first time the Holy See has taken a position on AI. In May, Leo XIV presented his first major doctrinal text, the encyclical "Magnifica Humanitas," focused on the risks of automation replacing human discernment in key decisions. In the same September 16 audience, the Pope linked his warning about algorithms to a condemnation of every form of attack on life —homicide, genocide, abortion, euthanasia and voluntary suicide— in what amounts to a single underlying argument: there are decisions that, for the Vatican, belong inalienably to the person and admit no delegation, neither to another power nor to a machine.

In general, the warning comes at a moment when the algorithmic delegation of personal decisions is no longer a hypothesis: chatbot-based companion apps, assistants that suggest who to date or what to reply, and recommendation systems that shape social consumption are an everyday part of how millions of people relate to one another. Governments, medical associations and educators have spent months debating, with varying degrees of evidence, the effect of these tools on isolation and emotional development, especially among minors.

What is relevant about this intervention is not the diagnosis itself —concern about screen-mediated loneliness predates generative AI— but that the planet's most organized moral authority is now framing it in explicitly algorithmic terms: the problem is not that the machine converses, but that we normalize handing over our judgment to it. It is a distinction that connects with something we have been observing in our sector-by-sector analysis of employment: where relationship, judgment and presence matter, AI performs worse and humans hold up better; the risk is not technological, it is one of design and habit, when the convenient shortcut replaces the effort of relating instead of supporting it.

In the short term, that is the real friction: products optimized to maximize engagement can push the most vulnerable profiles toward dependency, and no adoption figure for companion chatbots offsets that cost if it erodes real bonds. In the long term, however, none of this contradicts the underlying promise of this technology: an AI that frees up time, care and resources can strengthen community life instead of hollowing it out, as long as it substitutes for routine tasks and not for the bond itself. Leo XIV's warning, more than a brake, is a useful red line for drawing that boundary while the industry keeps building the product.