AI: Artificial Intelligence, Real Human Cost

AI: Artificial Intelligence, Real Human Cost

Artificial intelligence captivates the imagination. Its promises of innovation in healthcare and education dominate the headlines. Yet, behind these hopes of  technological progress lies a largely hidden reality. Far from being magical and  autonomous, AI relies on a global production chain that is profoundly unequal and  largely invisible, sustaining the illusion of a clean technology. 

I. The Myth of Autonomous AI 

Behind the apparent autonomy of artificial intelligences, a vast workforce labours daily to  supply systems with data.According to the World Bank, nearly 430 million moderators, often  in precarious situations, are tasked with training algorithms by classifying content viewed in  rapid succession. This repetitive work involves teaching machines to recognise images of  cats, dogs—in other words, to perceive the world through millions of human clicks. Without  this practice, machines would be incapable of identifying, let alone interpreting, human  behaviour, and thus unable to imitate it. Their tasks, though essential, are often harrowing,  requiring them to process randomly assigned shocking, violent, and toxic content such as  scenes of violence, photographs of corpses, or even acts of paedophilia. They must classify  these according to categories like “sexual,” “violent,” or “self-harm.” 

These “data workers” form the backbone of AI development: it is estimated that 80% of  the time devoted to an AI development project is dedicated to data collection,  preparation, and annotation (Cognilytica Research, 2019). In 2022, the global market for  data collection and labelling was valued at $2.22 billion and could reach $17.10 billion  by 2030 (Grand View Research, 2022). To meet this growing demand, major companies  outsource these tasks via Business Process Outsourcing (BPO) firms such as Sama, or  through digital platforms like Amazon Mechanical Turk, which assign tasks to freelance  workers. Under the guise of an ostensibly ethical model promoting the inclusion of  marginalised populations, these arrangements conceal degrading working conditions,  while allowing client companies to legally absolve themselves of any responsibility. But  behind this invisibility of human labour lies a brutal reality: a global system of  exploitation, masked by the promise of instantaneous, e\ortless technology. 

II. An Economy Built on Silence and Forgetting 

Henry Poulain’s documentary, The Sacrificed of AI, exposes a digital industry built on the  silent exploitation of thousands of workers. Subcontracting companies target regions of the  world afflicted by poverty and economic crises to recruit low-cost labour. In Kenya, “data  workers” earn between $1.34 and $2 per hour—a subsistence wage that aligns with criteria  for economic dependency (Casilli, 2024). 

In order to survive, these workers are compelled to accept inhumane conditions: night shifts,  unsanitary environments, complete lack of social protection, and above all, repeated exposure  to traumatic images. According to Maunick (2025), 90% of the content processed by these  moderators comprises images of extreme violence, corpses, acts of torture, or rape.The  psychological consequences are severe: anxiety disorders, depression, post-traumatic 

stress. These sufferings, though very real, are systematically ignored and denied by client  companies. The workers themselves speak of “mental torture” (Maunick, 2025). 

This distress is exacerbated by enforced isolation. Strict confidentiality clauses prohibit any  discussion of their work, even with close relatives, under threat of severe legal  action. Deprived of support and cut off from social contact, these workers remain invisible,  unable to organise collectively or join trade unions. This silence is no accident—it is  carefully orchestrated and sustained to uphold a global system of exploitation. In this way, AI  systems reproduce not only individual but also structural and societal forms of alienation. 

This outsourcing model allows major tech companies to abdicate responsibility. By avoiding  direct employment, they sidestep legal obligations and accountability. These tasks are framed  as minor execution work, deliberately downplaying their strategic importance. Yet without  these workers, AI systems could not function. It is they who provide the coherence and  relevance behind what we call “intelligent” models. 

III. Justifying Human Sacrifice 

To conceal the systemic exploitation they oversee, tech giants deploy ideological narratives  designed to legitimise the suffering built into their business model. Several researchers,  including Antonio Casilli, interpret this as a continuation of colonial logic: Western  technology firms outsource the most arduous and degrading tasks to the world’s most  vulnerable countries, where poverty traps local populations into accepting survival wages.  “They profit from our poverty,” says a worker in The Sacrificed of AI. Offering employment  under such degrading conditions is not a mark of progress but a veiled form of economic  domination. This shadow economy, outsourced to fragile nations, reveals a striking paradox:  while artificial intelligence is held up as a symbol of twenty-first-century innovation, its  operations still rely on subcontracting models inherited from colonial systems—where human  costs are deliberately overlooked. In doing so, it perpetuates a profoundly unequal global  system, in which the rewards are captured by a technological elite. 

To justify this exploitation, companies claim that artificial intelligence will eventually  become self-sufficient through machine learning, ultimately rendering the work of “data  workers” obsolete. This promise is rooted in a belief system known as longtermism, which  asserts that present-day suffering is the necessary cost of a utopian future. Casilli (2025)  highlights how this doctrine—popularised in the English-speaking world—elevates the  pursuit of a better future to a kind of “moral priority,” disregarding the real and immediate  harm it causes today. 

This logic is precisely what Sam Altman, CEO of OpenAI, endorses in his vision of a  technological paradise: “Robots will steal jobs and kill people, but on the whole, the outcome  will be extremely positive for humanity” (Poulain, 2024). This ideal draws on a  transhumanist ideology that champions the enhancement of humanity through technology for  the benefit of future generations. Such a worldview conveniently downplays the human and  social costs of AI, justifying the alienation, marginalisation, and exploitation of thousands of  workers. Artificial intelligence cannot be deemed ethical if it relies on immoral labour  practices.

Conclusion 

Today, artificial intelligence—hailed as a major leap forward—is advancing at the cost of  invisible human sacrifices. Behind the illusion of instantaneity lies a global production chain  defined by exploitation, precarity, and silence. It is essential to look beyond the interface and  consider the full extent of this often-overlooked supply chain. 

As users, citizens, researchers, or digital professionals, we all share a responsibility: to  understand what makes these technologies possible, to interrogate the conditions under which  they are produced, and to demand greater transparency, social justice, and ethical  accountability. Artificial intelligence can only be a genuine form of progress if it is built upon  a foundation that respects the lives it draws upon. 

References: 

• Aleteia. (2025). In the Shadow of AI: The Sacrificed of Longtermism. 

• France Info. (2024). “They profit from our poverty”: Behind the Boom in Generative  Artificial Intelligence, the Hidden Labour of AI’s Invisible Workforce. 

• L’Express Maurice. (2024). Who Are the Data Workers? The Sacrificed of Artificial  Intelligence. 

• Radio France – France Culture. (2024). Artificial Intelligence: During the Paris Summit,  Millions of Sacrificed AI Workers Were Forgotten. 

• Stereolux. (2024). Invisible Workers: The Hidden Face of AI. 

• Libre à lire. (2024). The Sacrificed of AI: The Overlooked Scandal of Data Workers. 

• Weizenbaum Institute. (2024). Data Workers: The Working Conditions and Significance of  the People Behind AI.