Good evening, reader.

Imagine, for a moment, an old man at a desk in late autumn in Rome. He is wearing a white cassock. The desk is older than most countries. In front of him sits a manuscript that has been worked over for the better part of a year by Jesuits, Dominicans, a small group of moral theologians, two retired judges, and at least one engineer. It runs to roughly 82 pages in the final English translation and a little over 42,000 words. He signs it on 15 May 2026. The date is not accidental. Exactly 135 years earlier, in the same city, his predecessor of the same regnal name signed a different document, Rerum Novarum, into the worst social wreckage of the first industrial revolution.

On 25 May 2026, the Holy See presented Magnifica Humanitas, the first encyclical of Leo XIV. It runs to roughly 82 pages and around 42,300 words in English, with a press-room note confirming it was signed on 15 May 2026, the 135th anniversary of Leo XIII's Rerum Novarum. The subtitle is operative: "On safeguarding the human person in the time of artificial intelligence."

It looks like a religion story. It is filed in newspapers next to stories about cardinals and conclaves and processions. It is written in a register that uses words like "dignity" and "person" without quotation marks. It does not appear in any of the registers that AI policy is supposed to appear in — not in the Brussels comitology, not in the West Wing, not in the AI Office's working groups, not in the UN's Independent International Scientific Panel. Nobody on the standards track commissioned it. Nobody on the standards track approved it. Nobody on the standards track can revise it.

That is the feature. It is not the bug.

The document is a governance intervention that operates one level above the texts written by lawmakers, regulators and standards bodies. It does not regulate AI systems. It regulates the frame in which AI systems are regulated.

Magnifica Humanitas refuses the existing frame. The encyclical's central instrument is a substantive theory of the person (dignity rooted in existence rather than capacity, output or function) against which the claims of AI systems are to be measured. The document then derives obligations on developers, deployers, governments and users from that anthropological floor. The order of operations matters. In the standards-body frame, governance follows from risk. In the encyclical's frame, governance follows from a prior account of what a human is and is owed.

That order of operations is what makes it dangerous to incumbents, and useful to anyone trying to design honest governance.

What "disarm" means

The verb in the speech is unusual. Speaking to a Synod Hall packed with bishops, journalists and a handful of bemused diplomats on the morning of 25 May, Leo XIV said that "artificial intelligence needs to be disarmed... freed from logics that turn it into an instrument of domination, exclusion, and death. Like nuclear energy, it must be at the service of all and of the common good."

The analogy is the clue. Disarmament, as a concept, was almost entirely the property of one branch of governance for the second half of the twentieth century: nuclear weapons treaties. The Non-Proliferation Treaty of 1968, the Strategic Arms Limitation Talks, the Intermediate-Range Nuclear Forces Treaty, the test-ban regimes. Disarmament is what you do with capabilities that are too consequential to leave to commercial competition alone. It is treaty work. It is dual-use export-control work. It is "this technology is so dangerous that the question of who owns it cannot be settled by who can build it first" work.

Compare this with the verb the EU AI Act uses, which is "classify." Or the verb the Trump executive order of 11 December 2025 uses, which, as Sidley's analysis notes, is "preempt." Or the verb the UN Global Digital Compact uses, which is "coordinate." Each verb is the policy, and what each regime thinks its for.

What 1891 was about, and why 2026 rhymes with it

Rerum Novarum is the founding text of modern Catholic social teaching. Leo XIII issued it in May 1891, into the wreckage of the first industrial revolution, when European cities were filling with men whose hands had been shaped for one trade and whose trades had been broken by another, and when the prevailing answers from the political economists of the day were either that this was a regrettable transitional cost, or that it was so intolerable that only revolution could fix it.

Rerum Novarum refused both answers. It said something that is now so embedded in the labour law of every Western democracy that it is hard to remember it was once contested: that the worker is a person before he is an input; that a wage is just only if it allows a person to live decently; that workers may freely associate; that the state has a positive duty to protect the weak from the strong. The encyclical fed into the early-twentieth-century industrial relations settlements of most of the countries that were industrialising at the time. In Australia, it fed quite directly into the reasoning of Henry Bournes Higgins, the judge who, in his Harvester judgment of November 1907, looked at H. V. McKay's harvester factory and asked not what the market rate for a labourer was, but what wage a labourer needed to support himself and his family "as a human being living in a civilised community." From that small clause came the legislated minimum wage. From the minimum wage came the rest.

The successive Popes who commemorated Rerum Novarum on its anniversaries (Pius XI in 1931 with Quadragesimo Anno, John XXIII in 1961 with Mater et Magistra, John Paul II in 1991 with Centesimus Annus) each restated the doctrine for a transformed economy. Each was treated, in its moment, as a churchly artefact. Each turned out, in the longer run, to be a frame-setting document that subsequent legislation eventually had to honour, because once a serious moral language exists for a problem, the people writing the legal language eventually have to reckon with it.

Leo XIV chose the same anniversary on purpose. He chose the same regnal name on purpose. The argument is openly stated: this is the technological era's labour question. The hands being broken this time belong to writers, paralegals, junior coders, radiologists, translators, customer-service workers, illustrators. The factory floor is the corporate workflow. The political economists of the moment are saying, again, that this is a regrettable transitional cost. Or that the labour question is solved, this time, by waiting for capital allocation to do its work.

Magnifica Humanitas refuses both answers, as its grandfather refused both answers. Its instrument is the same: a substantive theory of the person. Dignity is not earned by productive capacity. It is not granted by the regulatory state. It is owed because of what a human being is. And from that floor, the rest of the obligations on developers, deployers, governments and users are derived.

This is the structural move. It supplies what every existing AI governance regime quietly lacks: an external normative anchor that incumbents cannot capture by sending lawyers to working groups. The encyclical does not have legal force. It does not need to. It has frame-setting force, which is a more durable resource and one that compounds across multiple legislative cycles.

What the standards track is presently doing

It is worth pausing to be precise about what the three live AI governance tracks are, and what they are not.

The EU AI Act is a risk-classification regime. The largest model providers: OpenAI, Anthropic, Google, Meta, Mistral, xAI, engage with the regime through the GPAI Code of Practice and structured access requests from the AI Office. The Act's central question is: which capabilities, deployed where, at what risk level. The frame is technical conformity.

The Trump executive order of 11 December 2025 is the opposite move on the same axis. It instructs the Department of Justice to identify and litigate against state AI laws deemed "onerous," conditions federal grants on state compliance, and directs the Federal Communications Commission and Federal Trade Commission to preempt conflicting state rules. The stated policy is "to sustain and enhance the United States' global AI dominance through a minimally burdensome national policy framework for AI." The Act's central question is: which level of government owns the rulebook. The frame is jurisdictional consolidation in the service of competitive position.

The UN Global Digital Compact is a third track, multilateral, slower, oriented towards capacity-building. Its Independent International Scientific Panel reports from 2027. The frame is coordination.

These three tracks disagree about almost everything except the frame. All three accept that AI governance is, at root, about classifying capabilities, allocating jurisdictional authority, and managing the rate at which capabilities reach markets. The encyclical is doing something else. It is asking what is being classified, by whom and for whose benefit, and whether the act of classification itself can carry the freight that has been loaded onto it.

Its answer, from the text itself: "technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it. Therefore, the primary choice is not between a 'yes' or 'no' to technology, but rather between constructing Babel or rebuilding Jerusalem." Translated out of the metaphor, the claim is that no risk-classification regime is sufficient on its own, because the regime itself is shaped by the interests that fund and operate it. This is closer to the academic literature on regulatory capture than to anything published by a faith institution in recent decades. It is also, awkwardly for the technocratic mood, correct.

What Australia does with it

Australia's current AI governance posture is a Voluntary AI Safety Standard published in September 2024, ten guardrails covering accountability, risk management, data governance, testing, human oversight, transparency, contestability, supply-chain compliance, record-keeping and stakeholder engagement, plus a proposed mandatory guardrail regime for high-risk settings that has not yet legislated. The Productivity Commission's interim AI report and the Department of Industry's consultation on mandatory guardrails sit within the same frame as the EU and US debates: risk classification, capability assessment, conformity.

There is no Australian equivalent of an encyclical. There is no normative anchor written from outside the standards track. The Voluntary AI Safety Standard's preamble speaks of "shared values" and "social licence" but it does not articulate a substantive account of the person against which AI claims can be measured. It cannot, because it is itself a product of the standards track.

This is not a complaint about the Standard. It is well-drafted within its frame. The complaint is that the frame is incomplete. Australia is a country with a particular history of writing labour law against the grain of imported orthodoxy: the Harvester judgment of 1907 produced the world's first legislated living wage and did so by anchoring the calculation in "the normal needs of the average employee, regarded as a human being living in a civilised community." That phrase is doing the same work as Magnifica Humanitas's "dignity of existence", it is a normative anchor outside the technical machinery of wage-setting that nonetheless governs how the machinery operates.

A serious Australian engagement with the encyclical would not be ecclesial. It would be jurisprudential. It would ask which existing Australian normative anchors (the Harvester reasoning, the Mabo doctrine of inherent rights, the constitutional implications around political communication) can be put to work against AI claims in the way the encyclical puts dignity to work. It would treat the encyclical as a worked example of how a non-regulator can produce a frame that regulators must subsequently honour.

The country that wrote the first living wage is unusually well-placed to write the first living-with-AI standard. Not because it has the largest AI sector, but because it has the most mature tradition of refusing the frame in which living standards become a residual category after capital allocation. That tradition is currently dormant in the AI policy debate. Magnifica Humanitas is a reminder that it does not have to be.

Frugal comfort

There is a line near the end of Higgins's Harvester judgment that almost nobody quotes. He has done the arithmetic of the seven shillings. He has set out his reasoning about the human being living in a civilised community. And then, in passing, he writes that the measure of a wage is whether it allows the worker's home to attain a state of "frugal comfort."

Frugal comfort. It is one of those phrases that does not survive into modern policy writing. It is too quiet. It does not optimise anything. It refuses to be a metric. You cannot put it in a regulation. You can, with some effort, put it in a piece of reasoning that a regulation must then honour.

The encyclical, 119 years later and in another language entirely, is trying to put a similar phrase back into the room. Not the same phrase. The vocabulary is different and the tradition is different and the country is different. But the move is in the same spirit. There is an ordinary human, in an ordinary room, doing ordinary work that is now being done partly by a model trained on the work of other ordinary humans in other ordinary rooms, and the question of what they are owed is upstream of any regulation that classifies the model or pre-empts the state law or coordinates the international response. The question is what they are. The frame is what they are. The standards track does not produce that frame. It receives it, eventually, from somewhere else.

In 1907 the somewhere else was a Melbourne courtroom and an Irish-born judge who took a small statute and asked a large question. In 2026 it is an old man at a desk in Rome, with a Latin manuscript and a 135-year-old anniversary. In 2030, or 2035, or some date soon, the somewhere else will be someone else or some combination of voices working in a register that is not yet quite legible to the policy press. The frame-setting documents are not written where the frame is being argued about. They are written where someone has the time and the standing and the unfashionable courage to ask the older question.

The Innovation Commons is interested in who, in this country, is preparing to write that document. Not because the encyclical needs an Australian echo. It does not. Because the country that produced Higgins has the tradition, and the language, and the legal imagination to produce a serious answer of its own. Frugal comfort, in 2026, would be a good place to start.