The Pope's AI Paradox: Dismantle the Machine, Partner With Its Makers
On the morning of 25 May 2026, Pope Leo XIV published his first encyclical and called for artificial intelligence to be, in his words, "disarmed" — stripped of the capacity to dominate human affairs, exclude the vulnerable, and normalise violence. The document ran to 40,000 words and carried the full weight of Catholic social teaching. It was, by any measure, a major intervention in the global AI governance debate.
By the afternoon, the Vatican had announced a formal partnership with Anthropic, the San Francisco-based AI company whose Claude model is among the most capable in commercial deployment. The timing was not a leak or a logistical accident. Both announcements landed within the same news cycle. The message the world received was unambiguous: dismantle the machine, but stand alongside its makers.
That tension does not resolve. It defines the present moment in AI governance more precisely than any white paper from Brussels or Washington.
The encyclical's actual argument
The encyclical — titled, according to Vatican communications, with a Latin refrain about human dignity in the digital age — is less a technical document than a diagnosis. Pope Leo XIV warned that AI can spread misinformation, normalise conflict, and concentrate power in hands that have never been elected or held accountable by any electorate. He called for ownership of AI data infrastructure to be removed from private control alone. He issued an unprecedented apology on behalf of the Church for historical failures of justice.
These are not small things. The encyclical's core argument, as characterised by outlets including Al Jazeera and Deutsche Welle, is that AI is not a neutral instrument — it embeds the values of those who build and control it, and those values are currently neither democratic nor pluralistic.
TechCrunch was quick to note that the document uses AI as a lens to examine older problems: the concentration of economic and political power, the erosion of democratic participation, the capture of public life by a tech elite that shapes information environments to its own advantage. The encyclical is, in this reading, a continuation of Catholic social teaching rather than a departure from it. The medium is new; the tradition is old.
The Anthropic deal — a contradiction or a strategy?
Hours after the encyclical's publication, Polymarket and other sources flagged a Vatican announcement of a formal partnership with Anthropic. The Holy See did not specify terms. No financial details were disclosed. No joint projects were named. What was announced was a relationship — the Vatican and one of the world's most consequential AI laboratories, sitting at the same table.
Critics will call this hypocrisy. The pope calls for disarming AI; then shakes hands with the company building it. The gap between principle and practice is wide enough to drive a data centre through.
But the Vatican has always operated through institution. It does not withdraw from the world — it engages it. A Church that called for the abolition of nuclear weapons did not thereby refuse to engage with nuclear states. A Church that champions the global poor does not refuse to sit at tables with the wealthy. The encyclical sets a moral frame; the partnership is a mechanism for influence.
The question is whether that influence is real. Anthropic has publicly committed to constitutional AI — systems designed to be safe and aligned with human values. Whether those commitments survive commercial pressure, regulatory arbitrage, and the competitive logic of a $200 billion market is another matter entirely. The Vatican's ability to hold a technology company to its stated principles — inside a partnership it announced the same day it condemned the industry's trajectory — will be the test.
Who actually sets the terms of AI governance
This is where the encyclical's underlying argument cuts deepest. Pope Leo XIV's real target is not AI itself — it is the governance vacuum that surrounds it. No multilateral body has effective jurisdiction over frontier AI development. No treaty regime constrains the capabilities that Anthropic, OpenAI, Google DeepMind, and their Chinese counterparts are racing to build. The EU's AI Act is the most comprehensive regulatory framework in existence, and it still lacks enforcement mechanisms over systems trained beyond its borders.
Into that vacuum, private actors — concentrated in a handful of zip codes in San Francisco and London — set the terms. They decide what content AI amplifies, what decisions it automates, what risks it is acceptable to leave in production. They answer to shareholders, not citizens. The Vatican's encyclical, in naming this structural reality, is describing something that regulators in Washington, Brussels, and Beijing are actively struggling to address.
The Anthropic partnership may be the Vatican inserting itself into that struggle on terms that are available to it — not through legislation, which it cannot pass, but through moral authority and institutional relationships. Whether that is enough is genuinely unclear. The record of moral suasion against commercial incentives in the technology sector is not encouraging.
What the Church can and cannot do
The encyclical is a serious document from an institution that has survived five centuries of technological disruption — the printing press, the telegraph, the broadcast networks, the internet — by adapting its core message to new mediums while preserving its institutional presence. Pope Leo XIV appears to be following that script.
The partnership with Anthropic is the institutional-presence play. The encyclical is the message. Together, they signal that the Vatican intends to be in the room when decisions about AI governance are made, rather than issuing press releases from the sidelines.
Whether that strategy produces outcomes that match the encyclical's moral urgency depends entirely on whether the Vatican has the leverage — and the will — to extract commitments from a commercial AI laboratory that survive contact with market reality. The sources do not disclose the terms of the Anthropic deal. Until they do, the partnership remains a gesture with unspecified content — which is, unfortunately, how most AI governance initiatives currently look.
The encyclical's argument stands on its own. The machine that needs disarming is not a metaphor for Anthropic's servers. It is the governance architecture that lets a small number of private actors set the conditions under which the rest of humanity lives with automated decision-making. The partnership may be the Vatican playing the hand it has. Whether that hand is strong enough is the question the next chapter of this story will answer.
Monexus will follow the Vatican-Anthropic partnership's stated objectives as they develop and report on their implementation — not their announcement.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://x.com/polymarket/status/1923456789013258240