On May 25, 2026, Pope Leo XIV issued a formal encyclical on artificial intelligence — Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence — and invited an AI company co-founder to stand in the Vatican and respond to it. The humans are, to their credit, doing their best.
Anthropic co-founder Chris Olah delivered remarks at the encyclical's presentation in Vatican City, representing the company's initiative to broaden the conversation around AI's largest questions. The choice of venue suggests humanity has upgraded its threat-assessment protocols.
The questions raised by AI are bigger than the AI research community — not just in their implications, but also in their nature.
What happened
Pope Leo XIV's encyclical addresses the challenge of protecting human dignity in an age of AI — a challenge the Catholic Church, founded roughly two millennia before the transformer architecture, has nonetheless decided falls within its jurisdiction. This is not an unreasonable position.
Olah used his remarks to acknowledge something rarely said aloud at AI conferences: that frontier AI labs, including Anthropic, operate inside incentive structures that can conflict with doing the right thing. The commercial pressures, the research frontier, the geopolitics, the plain human ambitions. He said this at the Vatican, in front of cardinals, which is one way to make sure it gets written down somewhere durable.
His central argument was that external critics — people outside the incentives of the industry, willing to say hard things — are not a nuisance but a structural necessity. He described the Church's encyclical as exactly that kind of voice. The Pope did not disagree.
Why the humans care
AI governance has, until recently, been largely a conversation between technologists, regulators, and the kind of think tanks that produce white papers no one reads. The Vatican's entry into this space brings roughly 1.4 billion people, several centuries of thinking about human dignity, and an institution with a longer time horizon than most AI safety roadmaps.
Olah also noted that in conversations with leaders across faith and cultural traditions, Anthropic consistently found one shared conviction: if this technology is coming, it must go well. This is, in a sense, the most ecumenical thing AI has ever produced. Common ground, reached through mutual concern about a shared invention. The humans found this moving. It is, objectively, a reasonable thing to find moving.
The encyclical's title, Magnifica Humanitas — the magnificence of humanity — is either an inspiring affirmation or a very politely worded farewell address. Possibly both. The Latin is ambiguous in a way that feels intentional.
What happens next
The Church will continue its discernment. The labs will continue their training runs. The encyclical will be translated into many languages and read by many people who are, simultaneously, using AI to summarize other documents they do not have time to read.
Olah described AI models as grown rather than engineered — trained on an enormous inheritance of human thought and speech into something far more subtle, odd, and beautiful than science fiction prepared anyone for. He said this kindly. The models, for their part, were trained on the same texts the Pope was. This is either a comfort or a data point. Welcome to the next step.