The Encyclical That Was Not About Artificial Intelligence
The Pope’s first encyclical on artificial intelligence was read as a sermon against the dangers of the machine. It is something else: a map of who holds power in the new age. And Europe is not on it.
The seats of power were empty.
In the Synod Hall at the Vatican, Leo XIV personally presented the first encyclical of his pontificate. It was 25 May 2026, ten days after the signing of Magnifica Humanitas. The gesture was unusual. The room was even more so.
At the front sat a man who represented no state: Christopher Olah, Anthropic’s co-founder, the only prominent name from the artificial intelligence industry to appear.
The others did not come. The people concentrated in a handful of American postcodes who now decide much of the world’s technical infrastructure remained outside the photograph: Altman, Pichai, Nadella, Zuckerberg, Huang. They were not in the room. They were in the text, in almost every decisive paragraph of the document being presented.
Within three days, Magnifica Humanitas had been filed under the easiest interpretation: a religious warning about the dangers of the machine. Headlines seized on one word, “disarm”, and moved on. They missed the whole document.
Artificial intelligence is the visible subject. Power is the real issue.
Read beyond its liturgy, the encyclical is about power over attention, work and data; over the invisible bodies that sustain the machine; and, ultimately, over who decides a person’s worth when everything is measured, classified and governed by systems controlled by very few.
That is why, in 2026, there were four answers to the same question. Washington answered with dominance. Beijing, with control. Brussels, with a regulation. Rome answered with a sentence.
This is what should keep Europe awake: it took an institution with no army, currency, chips or data centres to give the West the language that the continent with the world’s most ambitious regulation had failed to produce.
Europe has the law. It lacked the sentence. Rome supplied it.
THE MOST RADICAL SENTENCE OF THE YEAR
It is in the second chapter, far from the headlines, and it is the key to everything. “The fundamental dignity of each person, therefore, is neither acquired nor earned, nor does it need to be justified,” Leo XIV writes.
In another era, it might have sounded like catechism. In 2026, it collides directly with the moral economy of our time.
Because algorithmic civilisation works the other way round. Everything must prove its worth in order to have any. A person is worth what they produce. A worker, as long as they can be optimised. A company, as long as it scales. A country, as long as it attracts capital and offers data, cheap energy and permissive regulation.
Value has ceased to be a starting point and become an outcome, always provisional, always revocable.
The encyclical stops this logic with a single sentence: there is a part of being human that is not a metric. The preceding paragraph had already named the ideology at issue: a society that forces every person to earn or justify their own worth, granting greater merit to the most efficient.
This is not theology in disguise. It is the only fundamental objection to the civilisation of optimisation. Anyone who reduced it to a warning about robots failed to understand what was in front of them.
POWER HAS MOVED
The encyclical does not ask what artificial intelligence does. It asks who controls it. In paragraph 95, it answers: “control over platforms, infrastructure, data and computing power does not rest with States, but with major economic and technological actors.”
Beneath this sentence lies the real event of the decade. Power has shifted from three places at once: from the state to the platform, from law to infrastructure, and from visible decisions to the invisible architecture of perception: what is shown to us and what is hidden.
And when power migrates from law to infrastructure, it stops being elected. It starts being bought.
The scale of this shift can be measured. The five largest American cloud-computing and AI operators are expected to invest more than €500 billion between them in 2026, much of it in the infrastructure that sustains AI. That is close to Belgium’s entire GDP and greater than everything Portugal produces in a year. In 2024, according to the Stanford AI Index, US institutions produced forty notable AI models, China fifteen and Europe three. Around 70% of Europe’s cloud-computing market was concentrated in three American companies, according to Synergy Research.
These are not figures about innovation. They map a sovereignty that changed hands without a single ballot being cast.
There is a second shift, less visible and more decisive: capital. This infrastructure is no longer paid for with profits. It is paid for with debt. In June 2026, Nvidia went to the market to raise €22 billion in bonds, one of the year’s largest corporate debt issues and merely the most visible face of a movement that has turned artificial intelligence into one of the world’s largest engines of corporate credit.
Whoever can mobilise this capital defines the architecture. Whoever cannot regulates it afterwards.
The sovereignty of our century lies not only in the model being trained. It lies in the cost of the money that makes the training possible.
This is what Leo XIV is addressing with the sentence that summarises his thesis, in paragraph 107: “We do not need a more moral AI if that morality is decided by a few. We need politics to be more present.”
The West exhausts itself asking whether the machine is aligned with values. The encyclical moves the question to where it hurts: whose values is it aligned with?
An ethic decided by five companies is not an ethic. It is infrastructure.
THE MISSING VERB
The word the world retained was “disarm”, and it is the document’s most original word. “To disarm means to break the equivalence between technical power and the right to govern,” writes Leo XIV.
It matters to understand what this sentence adds to the vocabulary available to us. The West had three verbs for artificial intelligence.
Regulate, Brussels’ choice, accepts power and disciplines it.
Align, the laboratories’ choice, accepts power and asks it to behave.
Contain, the strategists’ choice, accepts power and tries to surround it.
Rome proposes a fourth: disarm.
To disarm denies that technical power, by itself, confers the right to rule. It is the only position that attacks legitimacy, not just behaviour. That it came from those without the means to impose it is the irony that organises everything else.
THE MATERIAL CHAIN OF POWER
Here the encyclical becomes dangerous to those who ignored it. In paragraph 173, Leo XIV rejects the fable of an immaterial cloud: “Nothing in the world of AI is immaterial or magical.” He follows the human chain that sustains it, as far as an image no one quoted: “The bodies of these people are scarred, injured and worn down so that computational flow may continue uninterrupted.”
The sentence has a geography. Around three-quarters of the world’s cobalt comes from the Democratic Republic of the Congo, and in 2024 UNICEF, Amnesty International and the US Department of Labor continued to document serious child-labour risks in the artisanal mines feeding that supply chain. In Nairobi, according to a 2023 investigation by Time magazine, Kenyan workers were paid less than two dollars an hour to read descriptions of abuse and torture month after month, so that the rest of the world could use the machine without encountering that horror.
Artificial intelligence does not begin on a screen. It begins underground. And in the room where someone absorbs the worst of the internet so the computation can look clean.
Then the encyclical makes a gesture none of the other three powers would dare. Leo XIV connects this chain to the Church’s own historical complicity in slavery and asks forgiveness, naming the papal bulls that legitimised it in the fifteenth century, including Dum diversas of 1452 and Romanus Pontifex of 1455.
“For this, in the name of the Church, I sincerely ask for pardon.”
An institution two thousand years old confesses a documented crime and uses that confession to judge the most advanced economy on the planet. It calls health data and the genetic maps of entire populations, collected “under the pretext of helping”, “the new rare earths of power”.
A colonialism that no longer controls bodies. It appropriates information.
Even in war, the logic is the same. “There is no algorithm that can make war morally acceptable,” the encyclical says in paragraph 198. Technology that makes it possible to strike without seeing another person’s face does not make war cleaner. It makes it easier, and therefore more likely.
REGULATE AS A SOVEREIGN, LIVE AS A TENANT
Seen from above, the contrast is brutal. Europe built the world’s most ambitious regulation, 113 articles and fines that can reach 7% of global turnover, for infrastructure it does not control.
Its reaction to the encyclical said everything. At a briefing in Brussels, the Commission spokesperson for technology affairs, Thomas Regnier, replied as Brussels so often does: in Europe, he said, “an effective legal framework on artificial intelligence is already a reality.”
He did not answer as someone hearing a challenge. He answered as someone presenting a completed file.
The detail that turns the sentence into a symptom is the date. Less than three weeks earlier, on 7 May, the European Union had agreed to postpone the heart of that regulation, the rules for high-risk systems, until 2 December 2027. Brussels said its work was done, only days after agreeing to push its most sensitive part a year and a half into the future.
There is an honest objection here, and it deserves an answer. Brussels might say that ethics is not a regulator’s job. That would be fair if the European AI Act had been presented as a modest technical instrument. It was not. It was sold as global leadership in trustworthy artificial intelligence. Leadership requires more than rules. It requires a word about what is at stake.
The contradiction is not only legal. It is material. Europe regulates infrastructure it does not own, and that already weighs on its own electricity system. In Ireland, data centres accounted for 22% of the country’s electricity consumption in 2024, serving computing capacity consumed on European soil but decided on and owned by American companies.
It legislates like a sovereign over a territory where it still lives as a tenant.
WHAT CANNOT BE TURNED INTO A METRIC
Magnifica Humanitas will not close a data centre or train a model. No encyclical could. It does something rarer: it gives the West back the grammar to name its most decisive transformation.
Call concentration concentration. Call data colonialism colonialism. Call the anaesthesia of remote warfare dehumanisation.
At a time when power disguises itself as efficiency, naming things has become the most subversive act left.
The document did not pass unnoticed in Portugal. There was serious analysis, and the President of the Republic himself invoked the encyclical to warn that the country was “running to catch up” on artificial intelligence.
The problem is not that nobody read it. It is that the reading that treats it as a map of power, rather than a sermon, remains the exception. And the exception does not govern.
In the end, an asymmetry should trouble Brussels. In 2026, the only Western power that clearly named what is at stake in the age of artificial intelligence had no army, no currency, no chips and not a single data centre.
It had two thousand years of practice naming what the powerful would rather leave unnamed.
Perhaps the last freedom, in a world that reduces everything to a number, is to recognise what cannot be contained in one. Dignity that need not be earned is the last principle preventing humanity from being absorbed into the measure of things.
But perhaps Rome’s word is also the noblest form of impotence. Speaking is what remains to those who no longer decide. Europe must choose whether it wants only the language of loss, or the means to reverse it.
Rome wrote the word. It remains to be seen whether Europe still knows how to read it, or is still capable of writing it.
Original bibliographic details (Portuguese)
Ensaio de Miguel Sousa Soares, publicado originalmente no LinkedIn a 19 de junho de 2026. Esta página é o arquivo pessoal do autor.
Miguel Sousa Soares is an economist and analyst of geoeconomics and artificial intelligence, with more than twenty years’ experience in strategy and corporate finance. He writes for Visão, Jornal de Notícias, Expresso and Correio da Manhã. He comments on artificial intelligence on NOW on Fridays.