Human Dignity in the Time of Post-Capitalism
by Fr Stefano Giudici MCCJ
The Missionary Gaze
I would like to begin by telling you where I speak from, because it changes everything I will say afterwards. I speak from Nigrizia, from a gaze that we have learned to call “missionary” — which means, before anything else, a global gaze, one that starts from the last, from the impoverished. Looking from there changes the perspective. It also changes the understanding of what one is looking at. And yet we should humbly ask permission to be able to place ourselves where they are, to look from where they look: we do not look and speak in the name of the impoverished, but possibly with them. In fact, it should be they themselves who speak this evening.
AI and mission? Let us talk about artificial intelligence and work. To some it may seem a topic that has nothing to do with mission. I believe the opposite: it has become an unavoidable topic, and those who look from the impoverished and with them cannot afford to ignore it.
The Global South. We have heard what Magnifica Humanitas says about work, and in particular about a risk that the encyclical names clearly: the polarization between highly skilled workers and replaceable workers. My task this evening is to shift the gaze onto that same polarization, but seen from its most hidden root, in the Global South — where it begins, in the very way artificial intelligence is built, from an almost always invisible workforce. It is from there that I want to start again.
Three Ways in Which Artificial Intelligence Touches Work
First of all, let us distinguish, because it is not a single phenomenon. There are at least three different ways in which AI enters people’s work, and confusing them is the first, convenient way of not seeing what is really happening.
- The first is the work that produces artificial intelligence: those who train it, those who label its data, those who moderate what a model must never say. It is the most hidden level of all — often in Africa, often paid a few dollars a day. They are the ghost workers, and this will be the heart of what I tell you this evening.
- The second is the work that artificial intelligence organizes: delivery riders, drivers, gig workers, to whom an algorithm assigns the shift, evaluates performance, sets pay.
- The third is the work that artificial intelligence is transforming: those who write, program, analyze, translate, decide “using” artificial intelligence — and it is also here, to a large extent, that the polarization is at stake.
Three levels, three populations often on different continents — but a single thread binds them: who decides, who controls, who benefits, and who remains voiceless. I start from the first level, or at least I try. It is the most hidden, the most invisible.
The Real Costs of Artificial Intelligence
Let us remain, then, at the first level: the work that produces artificial intelligence. What costs are we talking about, when we say that this work has a price that never appears on the receipt? There are at least three: the environmental impact, the impact on our autonomy of judgment, and the hidden human labor. On the first two, Magnifica Humanitas is clear — I cite them only, without dwelling on them because I believe they will be the theme of other meetings.
Broadening our perspective to the use of AI in society, we see that it is now embedded in decision-making processes across many sectors and at multiple levels: in communication, management and control. The gains in efficiency and the potential to improve certain services are clear, yet rapidly and uncritically adopting them exposes us to a range of risks, including the tendency to overlook the environmental impact. Current AI systems require enormous amounts of energy and water, significantly influencing carbon dioxide emissions, and place heavy demands on natural resources. As their complexity increases, especially in the case of large language models, the need for computing power and storage capacity grows too, which requires an extensive network of machines, cables, data centers and energy-intensive infrastructure. For this reason, it is essential to develop more sustainable technological solutions that reduce environmental impact and help protect our common home.1 (Magnifica Humanitas 101)
The third major challenge is intellectual and concerns knowledge. Without careful attention, an educational system lacking in a love for truth may emerge, in which an incessant flow of information replaces the essential exercise of research, reflection and discernment. As knowledge becomes increasingly fragmented, it becomes difficult to grasp reality as a whole, to ask profound questions about meaning, or to develop authentic, critical and creative thought. Many educators already report signs of dehumanization, where people may “know many things” but struggle to find direction in their lives, partly due to an inability to connect information with deeper knowledge or maintain a sense of purpose. A genuinely healthy attitude is needed, requiring rhythms that incorporate silence, in-depth study, reading and judicious analysis, for without these elements inner freedom may be compromised. (Magnifica Humanitas 146)
An Italian researcher has called this risk “the trap of cognitive delegation”: the real challenge is not technological, but anthropological, and it concerns our ability to remain thinking beings.2
The third cost, the one I would like to dwell on this evening, is the hidden human labor behind every answer that artificial intelligence gives back to us in an instant. The ghost workers, a term coined in 2019 by Mary L. Gray (anthropologist) and Siddharth Suri (computer scientist) to indicate work carried out by a human being but which the end client believes is done by an automatic process.3
Today, who makes language models, the artificial intelligence chatbots such as ChatGPT, Gemini, Claude, less toxic, less offensive, capable of refusing dangerous requests? It was not only the mathematics of algorithms. It is the fruit of submerged, indeed invisible, work called data labelling. It is the activity by which human beings read, one by one, texts, images or conversations, and classify them — toxic or non-toxic, dangerous or harmless, correct or incorrect — so that a model can “learn” from those labels to behave in a certain way. Behind every kind reply from a chatbot, there has been someone who spent hours teaching it, example after example, what not to say.
Between 2021 and 2022, OpenAI — also thanks to the resources obtained from Microsoft — relied on Sama, a company specializing in the annotation of training data for artificial intelligence (AI) algorithms, which hired Kenyan workers at less than two euros an hour to do exactly this: read, one after another, the most violent and disturbing content imaginable, and thus teach the model not to reproduce it.
This is how Alessandro Aresu recounts it — geopolitical analyst, scientific advisor to Limes, author of “Geopolitics of Artificial Intelligence” (2024), where he speaks of “artificial artificial intelligence,” according to the paradigm of Amazon’s Mechanical Turk.4
A couple of words of explanation, because Aresu’s play on words is precise, not merely brilliant. The “Mechanical Turk” was an eighteenth-century automaton that seemed to play chess by itself, and astonished the courts of Europe: in reality, inside, a human player was hidden. Amazon called “Mechanical Turk” — Turkish mechanical, precisely — its platform that distributes into micro-tasks paid a few cents work that appears automatic but is carried out by people of flesh and blood. “Artificial artificial intelligence” is therefore the expression, used by Jeff Bezos at the beginning, and which Aresu takes up for an intelligence that presents itself as automatic, as pure machine, but which on closer inspection hides, like the eighteenth-century automaton, real human labor — often poor, often invisible.
Those workers of the OpenAI contract had a name: one of them was Richard Mathenge, paid less than two dollars an hour to read and classify extreme content. And Sama, it must be said, was not new to this kind of work: years before signing with OpenAI, the same company had contracted the moderation of Facebook content. It is there, in 2019 — two years before the contract with OpenAI even began — that another moderator, Daniel Motaung, was fired for trying to organize his colleagues into a union. Motaung sued Meta and Sama before the Kenyan courts — not for symbolic compensation, but to have it recognized that behind a subcontract, in offloading the world’s most devastating content onto workers paid a few dollars a day, there nevertheless remains a responsibility: that of those who own the platform.
The legal battle has now been going on for years. In 2024 the Kenyan Court of Appeal ruled that Kenyan courts have the right to judge Meta — a first step. But the judgment on the merits, the one that should establish whether there has been human trafficking, unlawful dismissal, union repression, has been postponed several times: the last time on 12 February. Still today, while we are here talking about it, those workers are waiting. In the meantime, Sama has progressively closed its moderation activities in Kenya. “The continued delay in the hearing and decision of our petition is not simply a procedural setback: it is a profound failure of the judicial promise that ‘justice delayed is justice denied,'” as stated by Naftali Andati Wambalo, part of the group of people who filed the lawsuit against Meta.5
Yet, from that same affair, something resembling a response was also born: in May 2023, over 150 moderators who worked for the companies behind Facebook, TikTok and ChatGPT gathered in Nairobi and founded the first African union of content moderators. Motaung was there. Mathenge was there. They did not wait for someone to save them: they began to organize themselves.6
And it is not an isolated case. After the Kenyan scandal, Meta simply moved much of its moderation elsewhere: to Accra, in Ghana, where another subcontracting company, Teleperformance, employs about 150 moderators, almost all from East Africa, with temporary work visas. An investigation published in April 2025 by the Bureau of Investigative Journalism recounted the same dynamics, aggravated: base salaries of just over 80 dollars a month — less than a fifth of the cost of living in Accra —, shifts in which one has as little as twenty seconds to decide whether a piece of content should be removed. One of the moderators, whom the investigation calls by the pseudonym Solomon, attempted suicide in December 2024, was hospitalized in a psychiatric hospital, and was eventually fired and repatriated, with just two months’ severance pay. Another, Abel, fired at the end of January on charges of “unauthorized absences,” had been fighting for better working conditions. “It is the worst conditions I have seen in six years of working with content moderators,” stated Martha Dark, of Foxglove — a British non-profit organization that uses legal action and public campaigns to make technology fairer — who has been following these affairs for years.7
Nairobi is not an exception: it is a model, one that repeats itself
And here it is, that same polarization just spoken of, embodied in a single company. Aresu himself recounts that the supercomputer Microsoft built in Iowa to train these models involves barely three hundred full-time employees — but they are, by his own specification, “laborers, electricians, technicians,” that is, personnel employed in the construction of the data centers, not in their management nor in the design of the models: the same category of manual and barely visible labor we are talking about this evening, this time in the heart of rural America, not in Africa. Another figure, concerning entirely different people in the same company, must be kept well distinct from this one: according to Bloomberg, the computer engineers and researchers of OpenAI — employed mostly in Menlo Park, California — can earn up to 800,000 dollars a year — an enormous figure (and yet equal to little more than a single day of what a footballer like Cristiano Ronaldo earns on the field, about 300 million dollars a year according to Forbes8). They are not, then, the same three hundred workers of Iowa: they are two opposite ends of the same supply chain, separated by a salary abyss while remaining in the same country. It is the polarization described by Magnifica Humanitas9, and just illustrated by Raffaella, made flesh and figures. Less than two euros an hour, in a supply chain that generates billions of dollars in valuation and that, at the other end, pays up to 800,000 dollars a year: no further comment is needed.
A Gaze That Widens: The Structural Causes
Before widening the gaze, a point of method: Richard Mathenge, Daniel Motaung, Solomon, Abel are not examples or statistics. They are people, with a name and a dignity that no analysis can put in second place. Keeping them present by name, to the end, is already a way of doing them justice.
Precisely out of fidelity to them, the Church’s social doctrine asks us for one step more: not to stop at the individual case, but to seek its causes — because an injustice that repeats itself identically, from Nairobi to Accra, is no longer merely the fault of a bad manager, but the sign of something structural. Magnifica Humanitas uses here the language that social doctrine has carried forward since John Paul II, that of “structures of sin“:
The idea of “social justice” helps us recognize that injustices do not arise solely from the wrong choices of individuals, but also from structures, mechanisms and economic and cultural systems that produce inequality almost automatically. Saint John Paul II spoke in this vein of structures of sin that oppose God’s will and require a commitment to personal and social conversion. In this perspective, justice is not merely about the fairer distribution of resources or the correction of current injustices, but also assumes a restorative dimension. It aims to mend broken bonds and reintegrate those who have been excluded, taking into account the wounds caused by injustices, such as wars, colonialism, racial or gender discrimination, violence against entire peoples and exploitation. (Magnificas Humanitas 79)
It is the same logic that the Church has applied, from Paul VI onward, to every development that leaves entire peoples behind (MH § 36, which I leave below for those who want the full text).
[A]s long as there are people in the world who are excluded from the development befitting human dignity, the Christian community cannot be content with a theoretical proclamation of peace. Rather, beginning where people are marginalized, it must allow the Gospel to pass judgment on those economic and political structures which — as John Paul II would later remind us — can become veritable “structures of sin.” As a result, no person or people will be treated as expendable in the processes of development. (Magnificas Humanitas 36)
Seeking the causes, then, and thinking of a social transformation that removes them, is not a way of minimizing the exploitation of Mathenge, of Motaung, of Solomon: it is, if anything, the only way not to leave them alone, each with his own case, while the structure that produced them continues intact to produce others.
And so we can say it: none of these phenomena — exploitation of labor, ecological disaster, taming of critical thought — is born with artificial intelligence. The older ones may perhaps remember the book “Letter to a Consumer of the North,” published in 1990. (Almost) exactly the same things were being said. It was the same injustice of underpaid and invisibilized workers. So these phenomena are, precisely, structural to the capitalist system, well before an algorithm existed. What is new, if anything, (but not even that much…) is the incredible and ever-growing capacity of this system to transform itself continuously, to absorb every critique and render it harmless. Indeed to recycle it and transform it into marketing.10
It is what Aresu again calls “the marketing of the concept of humanity.” The founding Charter of OpenAI, published in 2018, repeats the word “humanity” four times: it states that “our primary fiduciary duty is to humanity,” and that the development of artificial intelligence must never “harm humanity” nor “unduly concentrate power.” These are solemn statements — but no one has ever defined them in a verifiable way: what it means, in practice, to “harm humanity,” or what threshold distinguishes legitimate power from one “unduly concentrated,” no document says. It is precisely because of this vagueness, which allows humanity to be invoked as a justification while power over computing, data and customers is accumulated, that Aresu comes to write that one can speak, of this use of the term, as a “gigolo of humanity”: it is courted in words, without ever having to be held accountable in deeds.
It is precisely the question that Magnifica Humanitas poses that unmasks that rhetoric: it is not enough to declare oneself at the service of humanity, one must always ask
From this follows a simple but compelling consequence: we cannot consider AI to be morally neutral. In reality, every technical tool embodies choices and priorities through what it measures, ignores and optimizes, and how it classifies people and situations. If a system is designed or used in a way that treats some lives as less worthy, or excludes them without the possibility of appeal, then it is not merely a tool “to be used well,” since it has already introduced criteria that contradict the inalienable dignity of the human person. For this reason, ethical discernment cannot be limited to asking whether we are using a system for good or bad purposes; it must also examine how that system is designed and what vision of the human person and society is embedded in the data and models that guide it. (Magnifica Humanitas 104)
Why We Keep Speaking
At this point someone might object: if the system absorbs everything, if even critiques become marketing, what is the point of continuing to denounce? It is, at bottom, the oldest objection leveled at every prophetic voice: nothing will change anyway, so why speak?
I believe the answer does not lie in political analysis, but in the conviction — the heart of my missionary vocation — that denouncing an injustice remains a duty even without having, today, the power to correct it immediately. Political powerlessness is not moral powerlessness. To give up saying it because “it’s no use anyway” means handing over to the system even the last thing we have left: the word.
Can We Think of a Post-Capitalism?
And so, can we truly think of something different? I believe we can — not as a naive optimist, but as one who does not resign himself to the present as if it were the only possible world.
I am thinking of Rafael Díaz-Salazar’s Notebook, “Toward Post-Capitalism,” presented a few months ago here in Verona, at the Comboni Mother House: three commitments — radicalizing democracy, including economic and labor democracy; building ecosocialism; changing culture and lifestyles. Díaz-Salazar also said that artificial intelligence is “non-redeemable”: I do not quite agree. I believe it should be subjected to exactly that same radicalization of democracy, not condemned from the outset. I know it sounds almost utopian. But a true utopia is not a place that does not exist: it is a direction toward which to walk.
And it is not only direction: there are already signs, fragile but real.
Last 12 June, in Geneva, the ILO adopted — 406 votes in favor — the first international Convention on decent work in the platform economy: freedom of association, health and safety, remuneration, data protection, algorithmic management, the right to a review before the deactivation of an account.11
It must be said that the Convention speaks above all to the second level — riders, drivers — not yet to the first, that of Mathenge, of Motaung, of Solomon, of Abel. And yet the principle holds for them too: work mediated by a platform or by a subcontract does not leave the sphere of fundamental rights.
In 2023 the question was who the invisible people are who make AI work; in 2026 it is beginning to become legislation: what rights does one who works under the control of an algorithm have. They are the same question, asked twice.
And that union born in Nairobi in 2023, of Motaung and Mathenge, is still there demanding, together with a new association of data labelling workers. They are not yet the democratic State we imagine. But they are proof that the direction is not pure theory: it is already sprouting.
A Commitment on Several Levels
Our commitment, then, must move on several levels at once.
• Understanding the complexity of this reality, without yielding to the temptation to simplify it into black and white.
• Working so that every achievement of human research becomes the heritage of all, and not the property of a few.
• Supporting critical awareness and the real participation of peoples in the social and political life of their own territory.
• Giving space — certainly this is the task of Nigrizia — to critical and situated information on the side of the Global South, of those who suffer political capitalism, who do not govern it.
And I do not say this alone, nor only from a religious pulpit: the general secretary of CISL, Daniela Fumarola, said almost the same words, commenting on Magnifica Humanitas: mass education is “the great social divide of the digital transition” — a subjective right throughout one’s working life, not a privilege for a few. And, she adds, without participation “technology becomes technocracy.”1
A secular observation, which arrives at the same conclusion as the Church’s social doctrine by another road: here there is no room for sectarianism, only for the urgency of a common front.
I would like to close by returning to the three levels from which I started: it is there that the final judgment on everything I have told you is at stake.
I am not an apocalyptic: I do not much agree with those who consider artificial intelligence unredeemable. It can be placed at the service of the common good: healing, translating, lightening dangerous or repetitive work, opening knowledge to those who today remain excluded from it. This is its real positivity, not a naivety to be rejected.
But it remains unripe as long as it does not come to terms with what we have seen this evening: work made invisible — and, when someone tries to make it visible, like Motaung or the workers of Accra, it is often made even more invisible, with a dismissal, a repatriation, a contract terminated in six days. It is the question that every honest discourse on artificial intelligence must answer first, not last: who produces the intelligence, who suffers it, and which of the two we are truly able to see. And this, I repeat, is a structural problem of the capitalist system, not confined to the artificial intelligence sector.
We have been asked whether there are signs of an artificial intelligence that can enhance human dignity in mission contexts. This evening I answer yes, on one condition: the sign is not in the technology itself, but in the workers of Nairobi who founded a union; in an international convention signed despite all its limits. They are small signs, often uphill. But they are real. Continuing to recount them remains our way of guarding hope, without ceasing to name those who still have no voice.
1Aczel et al., Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints, UNU-INWEH, 2026: data centers could reach a consumption of 9.3 trillion liters of water and over 945 TWh of electricity by 2030 (6th place worldwide if they were a nation) — compared with the 5.5% of global emissions already attributed today to military apparatus alone. (Neimark et al., Queen Mary University of London, 2024)
2Cfr. Cosenza, Restare pensanti nell’era degli agenti, 2026.
3Gray M.L., Suri S., Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass, Houghton Mifflin Harcourt, 2019
4Alessandro Aresu, Geopolitica dell’intelligenza artificiale.
5Capital FM Kenya, “Delay in Meta Moderators Case Sparks Outcry from Petitioners”, febbraio 2026, https://capitalfm.africa/delay-in-meta-moderators-case-sparks-outcry-from-petitioners/
6Billy Perrigo, “150 African Workers for ChatGPT, TikTok and Facebook Vote to unionize at Landmark Nairobi Meeting”, TIME, 1° maggio 2023, https://time.com/6275995/chatgpt-facebook-african-workers-union/
7The Bureau of Investigative Journalism, “Suicide attempts, sackings and a vow of silence: Meta’s new moderators face worst conditions yet”, 27 aprile 2025.
8Forbes, Classifica mondiale degli atleti più pagati, ed. 2026.
9“Remunerazioni molto elevate per una minoranza altamente specializzata e salari sempre più ridotti per una larga parte della popolazione attiva”, Magnifica Humanitas, § 151
10Centro Nuovo Modello di Sviluppo, Lettera ad un consumatore del Nord, EMI (Editrice Missionaria Italiana), prima edizione 1990 (l’edizione poi ristampata, ISBN 9788830702769, è del 2004)
11International Labour Organization, Convenzione (n. 193) sul lavoro dignitoso nell’economia delle piattaforme, 12 giugno 2026, Conferenza Internazionale del Lavoro, 114ª sessione.
