The Hidden Human Cost of AI: What 'Magnifica Humanitas' Demands We See

Researchers and investigative journalists have documented for several years that the training of large language models depends on a vast, largely invisible workforce. These workers, concentrated in countries across sub-Saharan Africa, Southeast Asia, and Latin America, perform content moderation, data annotation, and the identification of harmful material so that AI systems can learn to avoid generating it. A 2023 investigation by TIME magazine found that Kenyan workers contracted through Sama to review disturbing content for OpenAI were paid less than two dollars per hour and reported severe psychological trauma from repeated exposure to graphic material. They described their work, in some testimonies, as a form of psychological injury for which no adequate support was provided (Perrigo, 2023).

Naftali Wambalo, a Kenyan father of two with a college degree in mathematics, told CBS's 60 Minutes what that work looked like from the inside: 'I looked at people being slaughtered.' He had taken the data-labeling job because it was available. The psychological cost was not part of the contract (Whitaker, 2023).

Pope Leo XIV's first encyclical, Magnifica Humanitas, released May 25, 2026, addresses this directly. The document warns that the efficiency gains AI promises cannot be celebrated if they are 'built on a chain of exploitation that remains deliberately hidden' (Leo XIV, 2026). Catholic theologian Léocadie Lushombo, speaking at the Vatican presentation before Cardinal Michael Czerny, called this inclusion prophetic. In her home country of the Democratic Republic of the Congo, she has witnessed children emerging from cobalt and nickel mines coated in toxic dust — the supply-chain labor that makes AI hardware possible and that almost never appears in conversations about artificial intelligence. She told the National Catholic Register that she encouraged people to 'care about knowing' how their use of AI products affects other human beings (Hackett, 2026).

That phrase deserves more weight than it initially carries. It is not a call for a boycott or a position paper. It is a call for a specific kind of attention — one the Catholic intellectual tradition has long understood as a moral capacity that can be trained or neglected, strengthened or atrophied.

What moral disengagement actually does

Albert Bandura's work on moral disengagement identifies two mechanisms that operate together in situations like this one: diffusion of responsibility across long supply chains, and the effective dehumanization of distant workers whose faces and names never appear in the product (Bandura, 1999). Both are present in the current AI economy. The labor is distributed across continents. The workers are absent from every marketing narrative that celebrates the intelligence of the systems they trained.

The Catholic Christian understanding of the person offers a direct correction to the assumption that the user of a tool stands outside the moral field that tool creates. The human being is a unified whole — intellect, will, emotion, body, and spirit — whose acts of choosing participate in the formation of the person who chooses. This is not a peripheral claim. It is the structural basis on which Catholic social teaching has insisted, across more than a century of documents, that the conditions under which people work are themselves moral realities (Pontifical Council for Justice and Peace, 2004).

Magnifica Humanitas is doing something specific when it names digital workers as a category of moral concern: it is performing re-engagement. It restores visibility to people whom the architecture of the global technology economy had rendered invisible. Proximity has never been, in Catholic moral reasoning, a prerequisite for obligation (Leo XIV, 2026).

Resilience is not the same as endurance

The workers described in AI labor reporting are not people who lack resilience. Many demonstrate extraordinary endurance under conditions that clinical frameworks would recognize as traumatic. What they frequently lack is access to what makes genuine resilience possible: stable employment, psychological support, wages that reflect the harm they absorb, and the basic recognition that their suffering is morally significant.

The therapeutic alliance — which clinical research consistently identifies as the strongest predictor of good outcomes in mental health care — works because it restores to a person the sense of being seen, of mattering, of existing in a relationship where their inner life counts (Norcross & Lambert, 2011). What Lushombo describes, at a societal scale, is the structural absence of that alliance for an entire category of workers. The encyclical's insistence on their dignity is, in effect, a demand that they be seen.

Simone Weil wrote that attention is the rarest and purest form of generosity — meaning something precise: that truly attending to another person, without projecting onto them or reducing them to a function they serve, is both extraordinarily difficult and extraordinarily valuable (Weil, 1977). Research on compassion fatigue confirms the point empirically: moral attention is not a fixed resource. It can be depleted, and it can be cultivated. Communities that practice deliberate attention to suffering through ritual, narrative, and communal reflection show greater capacity for sustained prosocial behavior than those relying on spontaneous emotional response (Figley, 2002).

Lushombo's call to 'care about knowing' is, in this frame, a call to a practice rather than a position. It asks people who use AI tools to seek out the knowledge that makes moral attention possible — to read the reporting, to ask the questions, to let the answers carry weight in subsequent decisions. This is the ordinary work of conscience operating in a context that did not exist a decade ago.

The person who closes their eyes to preventable suffering does not thereby protect their own wellbeing. They diminish it. That claim is native to the Catholic tradition and increasingly supported by the second-wave positive psychology of researchers like Paul Wong, whose existential approach treats moral responsibility not as a burden imposed on flourishing but as a constitutive element of it (Wong, 2011). Magnifica Humanitas is drawing on centuries of anthropological depth to make a point that the empirical literature is catching up to: the formation of persons who can see what their culture has taught them not to see is not peripheral to mental and spiritual health. It is part of what health actually requires (Leo XIV, 2026).

Three things you can do

Moral attention, as both the Catholic tradition and the psychological research suggest, becomes real through practice. Here are three concrete starting points:

• Read the reporting Seek out investigative accounts of AI labor conditions — the TIME investigation into Kenyan data workers, the CBS 60 Minutes segment, and similar journalism. Reading them, however, is not enough; the practice is to sit with what you have learned long enough for it to inform how you think about the tools you use daily.
• Ask questions of the companies whose products you use. Consumers, institutions, and parishes can write to AI companies and the organizations that license their tools to ask about labor standards, pay rates, and psychological support for data workers. Demand for transparency, even when it goes unanswered at first, shifts the conditions under which accountability becomes possible.
• Support advocacy organizations working on AI labor rights. Groups such as the Data Workers' Inquiry and the African Content Moderators' Union are working directly on fair wages, psychological care, and legal protections for the people whose labor underlies AI systems. Financial support, public solidarity, and institutional partnerships extend the reach of their work.

References

Bandura, A. (1999). Moral disengagement in the perpetration of inhumanities. Personality and Social Psychology Review, 3(3), 193–209. https://doi.org/10.1207/s15327957pspr0303_3

Figley, C. R. (Ed.). (2002). Treating compassion fatigue. Brunner-Routledge.

Hackett, C. (2026, May 27). At Vatican launch of Magnifica Humanitas, theologians call AI encyclical a moral turning point. National Catholic Register.

Leo XIV. (2026, May 25). Magnifica humanitas [Encyclical letter]. Dicastery for Communication, Holy See. https://www.vatican.va

Norcross, J. C., & Lambert, M. J. (2011). Psychotherapy relationships that work II. Psychotherapy, 48(1), 4–8. https://doi.org/10.1037/a0022180

Perrigo, B. (2023, January 18). Exclusive: OpenAI used Kenyan workers on less than $2 per hour to make ChatGPT less toxic. TIME. https://time.com/6247678/openai-chatgpt-kenya-workers/

Pontifical Council for Justice and Peace. (2004). Compendium of the social doctrine of the Church. Libreria Editrice Vaticana.

Weil, S. (1977). Reflections on the right use of school studies with a view to the love of God. In G. Panichas (Ed.), The Simone Weil reader (pp. 44–52). McKay. (Original work published 1951)

Whitaker, B. (2023, March 5). The workers who train AI are fighting for their rights [Television broadcast segment]. 60 Minutes. CBS News.

Wong, P. T. P. (2011). Positive psychology 2.0: Towards a balanced interactive model of the good life. Canadian Psychology, 52(2), 69–81. https://doi.org/10.1037/a0022511