That is why Leo’s argument is strongest when AI’s promise is taken seriously. The danger is not usefulness. The danger is usefulness without custody. The true choice is not speed or stagnation. It is whether speed remains answerable to the human being.
That answer is easier to write than to enforce. Custody is not a slogan; it requires institutions trusted enough to set limits, laws precise enough to survive contact with the technology, and publics patient enough to prefer accountable delay over frictionless surrender.
Even the Vatican presentation carried that tension. Christopher Olah, an Anthropic co-founder and interpretability researcher, appeared at the event, and Reuters reported that he argued AI must be guided from outside Big Tech.¹⁰ That was not a contradiction. It was the point. The builders were in the room, but the room was not theirs.
The same should be true everywhere else. In the newsroom, the room cannot belong to the image. In the classroom, it cannot belong to the shortcut. In the workplace, it cannot belong to the metric. In the court, it cannot belong to the risk score. In the war room, it cannot belong to the target list. The machine may assist judgment. It cannot be allowed to absorb it.
That is why the opening image works. The machine stands near the Pope, but it does not occupy the center. It is present, powerful, and impossible to ignore. It is also waiting. It has no conscience of its own, no memory of sin, no experience of mercy, no duty to the person its answer may affect. It can process the room. It cannot govern the room.
The responsible act may not look heroic. It may be a hand paused over the publish button. A teacher asking one more question. A judge refusing to let a score become a sentence. A manager remembering that productivity is not the same as dignity. A citizen waiting for evidence before sharing the image that confirms what he already wanted to believe.
That is the optimism in Leo’s encyclical. The machine has not ended the human story. It has made the old work harder to evade.
The answer is not to stop seeing, or stop inventing, or stop trusting every witness before the witness speaks. The answer is to slow the hand, trace the claim, and remember that speed is not truth.
The machine can calculate, imitate, accelerate, persuade, and assist. It cannot decide what a human life is for. It cannot carry the moral burden for those who own it, deploy it, obey it, or hide behind it.
That remains our work. And the first adult act in the AI room may be the smallest one: before the click, before the strike, before the score, before the answer becomes the decision, someone has to say, wait.
Bibliography
1. Associated Press, “Fake image of Pentagon explosion briefly sends jitters through stock market,” May 23, 2023; Reuters Fact Check, “Online posts reporting explosion near Pentagon on May 22, 2023 are false,” May 22, 2023.
2. Leo XIV, Magnifica Humanitas, May 15, 2026; Holy See Press Office, “Presentation of Pope Leo XIV’s Encyclical Letter Magnifica Humanitas,” May 18, 2026.
3. Reuters, “Quotes from Pope Leo’s document warning world of AI risks,” May 25, 2026.
4. Leo XIV, Magnifica Humanitas; Reuters, “Quotes from Pope Leo’s document warning world of AI risks.”
5. Reuters, “Quotes from Pope Leo’s document warning world of AI risks,” May 25, 2026.