Leo XIV disarms AI, Anthropic takes the podium

On May 25 at the Vatican's Synod Hall, Christopher Olah, Anthropic co-founder, atheist, 33 years old, stepped up to the podium next to Leo XIV. He came to hear the Pope say that artificial intelligence must be "disarmed, freed from the logics that turn it into an instrument of domination, exclusion and death." Four months earlier, two Holy See advisors appeared among the external contributors acknowledged in the "constitution" of Claude, Anthropic's flagship large language model, published in January.

That sequence is the right frame for reading Magnifica humanitas, Leo XIV's first encyclical, signed on May 15 and presented on May 25. 235 pages. The first pope in history to personally present an encyclical at a press conference. And incidentally, the sharpest act of political framing on AI published in 2026 by a non-regulatory institution.

A text tougher than the regulators

The Pope isn't doing ornamental theology. He names things. "Control over platforms, infrastructure, data and computing power does not rest with States, but with major economic and technological actors," he writes in paragraph 5. Later: "When such power is concentrated in the hands of a few, it tends to become opaque and evade public oversight."

The encyclical openly contests the dominant techno-libertarian narrative. "Change risks being governed solely by technocratic thinking, presented as necessary and inevitable," the text warns. Leo XIV then asks for very concrete things: independent audits, algorithmic transparency, equitable access to data, mechanisms of appeal. And the line that will stick: "It is not enough to invoke ethics in the abstract. Robust legal frameworks, independent oversight, informed users and a political system that does not abdicate its responsibility are required."

The signing date, May 15, 2026, is no accident: it is the 135th anniversary to the day of Rerum Novarum, the encyclical published by Leo XIII in 1891 in response to the industrial revolution and the abuses of early capitalism. The new pope did not pick the name Leo XIV by chance. He explicitly steps into the lineage of the pope who, a century and a half ago, forced Catholic doctrine to look squarely at the power imbalance between capital and labor. The implicit message: 2026 is to AI what 1891 was to the factory.

Six months of closed-door meetings

So much for the text. Now the production context, much less covered by the French-speaking press.

Since 2016, the Vatican has hosted the Minerva Dialogues, annual closed-door meetings on AI ethics at Rome's Santa Maria sopra Minerva basilica. Over the years, Eric Schmidt (former Google), Reid Hoffman (LinkedIn) have shown up. Sam Altman (OpenAI) and Demis Hassabis (Google DeepMind) have had private audiences with the Holy See.

In the six months leading up to Magnifica humanitas, engagement intensified. According to Religion News Service and Head Post, representatives from Meta, Google, Amazon, OpenAI, Anthropic, Palantir and Qualcomm met with Vatican officials. On April 29, 2026, a closed-door meeting took place at the French embassy to the Holy See, with tech executives, French officials and Vatican advisors. No public list of participants.

And then there's the Anthropic detail. In January 2026, the company published the new "constitution" of Claude, the document that codes the model's internal "morality." Among the external commenters acknowledged: Msgr. Paul Tighe, secretary of the Dicastery for Culture and Education at the Vatican, and Father Brendan McGuire, a former Silicon Valley engineer turned Holy See advisor. The same two advisors who internally helped shape the Pope's thinking on AI.

The mic handed to those the text criticizes

When May 25 came, the Vatican chose to share the Pope's podium with Christopher Olah, not with a precarious clickworker, not with an independent ethics researcher, not with a European regulator. Olah co-founded a company valued at several tens of billions of dollars, headed for an IPO. A company that belongs without ambiguity to the category of "major economic and technological actors" the text names.

Olah spoke politely. He said that "it is enormously important that there be people outside those economic incentives, who care about things going well, who insist on safety, who are willing to say hard things." The Pope, courteously, thanked him: "What a great sign of hope it is that with our differences we can listen to one another."

Several observers cited by Religion News Service noted that the sequence looked uncomfortably like a papal stamp of approval for Anthropic. A Vatican source tried to cut it short: Anthropic's presence "is not an endorsement, nor a prize, nor a reward, nor a canonization." The wording itself reveals what the Vatican feared the image would suggest.

The marginal cost of Roman soft-power

Then the awkward question. What does a company like Anthropic, Meta or Google come looking for in Rome?

Lobbying in Brussels around the EU AI Act runs into millions of euros per year per major player. Lobbying in Washington, into tens of millions. Lobbying in Rome means a few trips, a few dinners, a few consulting days given to the Curia. The return on investment, in image terms, has no equivalent. You get a global text, read on every continent, criticizing "the concentration of power" without naming a single company, and at whose launch you physically hold the microphone next to the spiritual leader of 1.4 billion Catholics.

The encyclical remains a text without binding force, deriving all its strength from symbol. And symbols, the targeted players know how to buy.

What the text does anyway

You can stop there and conclude that everything is performative. That would be moving a little fast.

First, Magnifica humanitas injects into public debate a framing that European and American regulators have not managed to impose: private power over AI is a civilizational problem, not a technical one. The text will be quoted, translated, picked up by NGOs, unions, parties. In the United States, where the Trump administration sanctioned Anthropic for refusing to license its technology for unrestricted military use, the text provides opponents with a moral reference. In Europe, where the EU AI Act is slowly entering its implementation phase, it strengthens the "power concentration" framing that the Commission was struggling to carry alone.

Second, the line that will stick from the encyclical names its own conditions of production: "A more moral AI is not enough if that morality is determined by a few." That applies to Anthropic's engineers, who wrote Claude's constitution after consulting two Vatican advisors.

It also applies to the Vatican itself, which drafted its encyclical after consulting major tech companies. The line turns on its authors. That may be what the text gets most right.