Disarming AI: What does it mean to "Disarm AI?"
Pope Leo XIV says AI should serve human flourishing. Meanwhile the Trump administration's data-center permitting order exempts AI datacenters from regulations designed to protect people.
The “American Pope” and the Trump Administration seem to be at odds on everything from detention center warehouses, to AI datacenters. The Pope wants to prevent AI from “Dominating Humanity” while the Trump Administration wants to prevent AI DataCenter buildouts from “unnecessary regulation.”
Pope Leo XIV’s encyclical was signed May 15, 2026 and publicly presented ten days later, it is his first encyclical, Magnifica Humanitas, says the work of governing AI is “to disarm” it — an expression, he writes, “which is close to my heart.” Disarming AI, he continues, “does not mean rejecting technology, but preventing it from dominating humanity. It means freeing technology from monopolistic control and opening it to discussion and debate, therefore making it human-friendly and restoring it to the plurality of human cultures and ways of life.” AI, in this frame, is subordinate to something: the flourishing of the humans who have to live inside whatever gets built.
This seems an almost direct response to document designed not to be seen, not to be noticed, and definitely not to be remembered, but every intention of being obeyed immediately: Executive Order 14318, “Accelerating Federal Permitting of Data Center Infrastructure,” signed July 23, 2025 — between the two dates above, and unconnected to either. It does not ask what AI is for. It answers a narrower, more operational question — what AI infrastructure is exempt from — and it answers that question in exhaustive, load-bearing detail: which environmental reviews don’t apply, which agencies must approve within ten days, which federal land can be leased, which coal plants must keep running to power it.
Put the two documents side by side and you don’t get a disagreement about AI policy. You get two entirely different genres of document, answering to two entirely different constituencies, and the distance between them is the whole story.
What the Pope subordinated AI to
Magnifica Humanitas is not, first, a document about artificial intelligence. It is a document about Catholic Social Doctrine — subsidiarity, the universal destination of goods, the ontological dignity of the human person — applied to a moment when, as Leo puts it, “the highest level is not the State, but rather major economic and technological actors that exercise de facto power over the conditions of everyday life.” AI is one case among several where that doctrine gets tested. But it’s the case Leo chose to name in the most pointed language in the whole document.
Three moves do the work.
Subsidiarity asks who is entitled to make decisions about the conditions of ordinary life. The doctrine’s classical answer is: the lowest competent level — the person, the family, the local community — with higher institutions existing to support that competence, not override it. Leo’s encyclical applies this directly to the platforms and infrastructure operators who have inverted the proper order, installing themselves as the de facto highest level of decision-making authority over daily life. The remedy he names is not abstract: “transparency, accountability and meaningful forms of participation (including independent checks, transparency regarding algorithms, equitable access to data and avenues for recourse).”
Ontological dignity asks what a person is worth, and answers: worth belonging to the person “simply by virtue of existing,” not by virtue of “what they achieve or produce.” Leo names the opposing anthropology explicitly — a “technocratic and post-humanist mentality that tends to regard the human person as an object to be manipulated or a resource to be optimized, removing all safeguards against the unchecked pursuit of profit.” That mentality, he writes, is where the AI-era harms he catalogs elsewhere in the encyclical — algorithmic decisions that “deny access to healthcare, employment or security based on unjust and prejudiced data,” the hidden labor of “data labelers” and rare-earth miners — actually originate. Not from the technology. From treating people as inputs.
The universal destination of goods asks who the value belongs to once it’s created, and extends the classical answer — everyone — to data itself: “ownership of data cannot be left solely in private hands but must be appropriately regulated. Data is the product of many contributors and should not be treated as something to be sold off or entrusted to a select few.”
None of this is anti-technology. Leo is explicit about that, in the same breath as the “disarm” language: the goal is not rejecting AI, but “preventing it from dominating humanity.” The frame throughout is subordination — AI answers to human flourishing, not the reverse. Doctrine, in this frame, exists to make sure that subordination holds; a government that takes it seriously would exist to enforce it.
What the executive order subordinated the public to
EO 14318 is what happens when a government writes the opposite kind of document — one where the subordination runs the other way.
Start with what it revokes. EO 14318 repeals an earlier Biden era EO 14141, “Advancing United States Leadership in Artificial Intelligence Infrastructure,” signed by President Biden on January 14, 2025 — six days before he left office. Whatever else can be said about that order’s follow-through, its text made public commitments. Its purpose section advances “leadership in the clean energy technologies needed to power the future economy,” commits to maintaining “low consumer electricity prices,” and states that “the development of AI infrastructure benefits the workers building it and communities near it” — commitments then formalized as five guiding principles the federal government would follow “as appropriate and consistent with applicable law.” Whatever their practical force, these are promises made to somebody — ratepayers, workers, host communities — named as parties the buildout owes something to.
EO 14318 makes no such promises.
There is no consumer-price commitment anywhere in its text. No worker-or-community-benefit clause. No preference for clean energy — its own definition of qualifying infrastructure explicitly includes “natural gas turbines, coal power equipment, nuclear power equipment, geothermal power equipment, and any other dispatchable baseload energy source,” with no ranking among them.
The public shows up in this document exactly once (structurally: as a variable in a siting calculation — land, water, and grid capacity to be allocated to a Qualifying Project) but ever as a party the government owes anything.
That absence isn’t incidental; it’s operationalized.
Section 5(c) of the order sets the rule that does the real work:
“Federal financial assistance representing less than 50 percent of total project costs shall be presumed not to constitute substantial Federal control and responsibility”
Meaning any datacenter project financed less than half by federal money is presumed, by definition, not to trigger the environmental review that “major Federal action” requires under the National Environmental Policy Act.
Hyperscale datacenters are financed overwhelmingly with private capital already, with federal loan guarantees and tax credits filling in a minority share. The 50% line isn’t a compromise; it’s a threshold engineered to sit above how these projects were already being financed — which means the buildout can take federal subsidy from Section 3’s loans, guarantees, grants and tax incentives, while remaining, for legal purposes, invisible to the public review process that subsidy would otherwise trigger.
That is the whole shift, in one provision: from a government promising to look out for the public inside the buildout, to a government engineering the exact proportion at which the public’s own money in the buildout stops counting as public.
Same technology, two different masters
Neither document disputes that AI infrastructure is going to get built at enormous scale. Neither is a document about whether. Both are documents about to whom AI answers while it gets built — and that is the question on which they are exact opposites.
Leo’s answer: AI answers to human flourishing, held there by doctrine that predates the technology by a century and is explicitly not negotiable to “efficiency” — his word for the thing ontological dignity refuses to be traded against. The government’s role, in his frame, is to enforce that subordination: transparency requirements, accountability mechanisms, “avenues for recourse” for the people the platforms have de facto authority over.
The executive order’s answer: AI infrastructure answers to its own buildout speed, and the public’s role is to not be in the way — to have its environmental-review rights presumed absent below a financing threshold calibrated to how the industry already spends money, to have its land and water treated as sitable resources, to have no textual claim on the clean energy, price protection, or community benefit the prior administration’s order at least wrote down as an aspiration.
You don’t need to be Catholic, or religious at all, to see what changed genre here. A text that names the public as a party owed something became a text that doesn’t mention the public except as terrain. That’s not a policy dispute about permitting timelines. It’s a decision about who an authority answers to — made within the same eighteen-month span, by a pope with no army and a president with the full machinery of a state, answering what looks like the same question and is actually two different ones: “is AI dangerous?” versus “is AI fast?”
Magnifica Humanitas never names Executive Order 14318.
It didn’t need to. A Vatican that tracks American AI policy closely enough to coordinate its own release with an Anthropic co-founder’s appearance at the encyclical’s Vatican unveiling was not writing in ignorance of what Washington was doing with the same technology.
Chris Olah stood beside the Pope that day, calling for AI labs to submit to “moral voices that the incentives cannot bend” — a plain acknowledgment that inside the industry itself, someone thought the incentives needed bending.
The permitting order is what it looks like when a state decides they don’t.
One quick ask.
A single Executive Order is a news story that slides into obscurity. But the patterns and connections are what drive my reporting.
At The RAMM I connect the dots that don’t fit in one article: the federal contracts, the 287(g) deals, the quiet real-estate moves, the county-by-county buildout of detention capacity.
The receipts:
4,800+ sourced events at CaptureCascade.org
~2000 counties with signals of detention expansion at detention-pipeline
150+ community fights over detention capacity, tracked
This is self-funded. Subscriptions are the only thing that keeps it alive