WHEN THE STATE SOLD CHILDREN: Switzerland's forgotten child-labour system—and what it reveals about labour exploitation in the age of AI
WHEN THE STATE SOLD CHILDREN
There are some historical photographs that become more disturbing the longer you look at them. Not because they show violence, but because nothing in them appears immediately violent. People are standing. People are working. Adults are going about their business. Institutions are functioning. The world looks orderly. And then you realise what the photograph does not show: the powerlessness of the child. The child has no voice, no bargaining power and, often, no meaningful way to refuse the arrangement into which adults have placed them.
In Switzerland, a country now associated with prosperity, precision, stability and one of the world's highest standards of living, vulnerable children were once turned into a source of labour. They became known as Verdingkinder, or “contract children”. Children were removed from families and placed with farmers and other households. Many were forced to work, and many experienced physical and psychological violence and sexual abuse. Switzerland's Federal Office of Justice now explicitly describes Verdingkinder as children forced to work on farms and exploited as cheap labour. The broader system of compulsory social measures and placements continued until 1981.
The phrase “child labour” is disturbing enough. But the Swiss story is more disturbing because it was not simply a story of employers secretly breaking the law. It was, in many cases, a story in which public authorities and social institutions became part of the machinery through which vulnerable children were removed from families and placed into environments where labour could be extracted from them. The historical reality was not one simple federal programme: cantonal and local authorities, guardianship structures and other institutions were involved. That makes the story more—not less—uncomfortable. Exploitation did not always arrive as a conspiracy. It could be embedded in ordinary administration. And that raises a question that reaches far beyond Switzerland.
The story begins with poverty. Switzerland in the nineteenth and early twentieth centuries was not the wealthy country the world knows today. Poor families could have many children and few resources. Agricultural households needed labour. Welfare systems were limited. Social attitudes toward poverty were often deeply moralistic. A poor family was not always seen simply as a family that needed assistance. Poverty could be interpreted as evidence of parental failure, moral weakness, social disorder or an inability to live according to accepted standards.
That distinction mattered. Because once poverty becomes a moral judgement rather than an economic condition, the state acquires a different kind of authority over poor people. It can decide what is best for them. It can decide whether parents are fit to raise their children. It can decide where children should live. And, eventually, it can decide where those children should work.
The result was a system of compulsory placements in which children from vulnerable families could be taken away and placed with other households. The Swiss authorities now acknowledge that hundreds of thousands of children and young adults were affected by compulsory social measures and placements up to 1981, with many suffering severe physical, psychological or sexual harm. The Verdingkinder were among them. They were children. But the system increasingly treated them as something else.
This is where the history becomes uncomfortable. The system could be presented as social assistance. A poor child would be taken into another household, given food and accommodation and put to work. On paper, this could sound like an arrangement designed to solve a social problem. But there was another side to the calculation. The child could work. The farmer needed labour. The public authority wanted to limit the cost of supporting the poor. And the child possessed almost none of the bargaining power that an adult worker could theoretically exercise.
That combination is dangerous. Because when the person performing the work has the least ability to refuse it, their vulnerability itself becomes economically valuable.
The historical record includes cases in which children were publicly auctioned or allocated according to economic considerations. But precision matters: this was not a simple federal system in which the Swiss government literally owned children and sold them nationwide. Local and cantonal authorities and other institutions were involved in placements, and the arrangements varied across time and place. The image is still almost too grotesque to accept: a child, a marketplace, an adult deciding whether to take the child, a system in which the economic value of the child's labour could become part of the calculation.
And yet the most important question is not whether the system looked cruel. It is why it could look normal to the people operating it.
Exploitation rarely introduces itself as exploitation. It arrives wearing respectable clothes.
Welfare. Discipline. Responsibility. Work ethic. Social order. Necessity.
The child is not being exploited, the argument can say. The child is being taught to work. The child is not being deprived. The child is being provided for. The child is not being used. The child is being given an opportunity.
That language matters because it changes what society sees. The farmer sees labour. The administrator sees a welfare case resolved. The taxpayer sees a cost reduced. The institution sees its mission being fulfilled. And the child sees something none of them necessarily have to see: a childhood disappearing.
The agricultural story is the most famous part of the Verdingkinder history, but Switzerland's history of child labour extended into industrial settings as well. A 2026 Swiss National Museum investigation describes a child labour institution established in the canton of Zug in 1855. It housed up to 100 children aged between 12 and 18 who worked in a nearby spinning mill, with half working during the day and half at night. The museum notes that children were valued as nimble, inexpensive labourers and were often put to work between and inside dangerous machinery.
Some worked under extraordinarily harsh conditions. The Swiss National Museum records working days of up to 16 hours in some Swiss child-labour settings. At the Zug institution, children received daily wages of 55 to 120 centimes, while 65 centimes was deducted for board and lodging; other charges could also be deducted. The numbers are almost beside the point. The structure is what matters.
The child worked. Someone else controlled the conditions. Someone else controlled the money. Someone else determined what counted as care. And someone else possessed the power to decide whether the arrangement continued.
It was not only the farm. It was a broader social architecture in which vulnerable children could become economically useful.
And that reveals another uncomfortable lesson. Child labour did not disappear simply because Switzerland became more prosperous. It required institutions, political pressure and regulation to change. The 1877 Swiss Factory Act eventually prohibited factory employment for children under 14 and imposed limits on working hours, representing a major shift in the state's willingness to intervene in the employment relationship. But protections did not immediately cover every sector. Agriculture and some small-scale industries remained outside the same restrictions.
This matters because it destroys another comforting assumption. We often imagine that economic development automatically produces better labour conditions. It doesn't. Economic development can produce enormous wealth while leaving vulnerable workers behind.
Institutions have to create it.
The deeper history of the Verdingkinder system is therefore not simply about cruelty. It is about power. Who gets to decide? Who gets to refuse? Who gets to negotiate? Who gets to leave? Who gets to complain? Who gets believed?
The farmer had choices. The administrator had choices. The institution had choices. The child had almost none.
That is the fundamental architecture of exploitation. Not simply low wages. Not simply long hours. Not simply bad conditions. The inability to say no.
Once that disappears, the economic relationship changes completely. The worker may still technically be “working”. But the bargaining relationship has been hollowed out.
And that is why the history of the Verdingkinder matters far beyond Switzerland. Because the most dangerous lesson is not that one society once treated children terribly. It is that a society can construct an economic system in which exploitation becomes rational to everyone except the person being exploited.
For decades, the history remained largely marginalised. The victims carried the memories. The institutions carried the records. The wider society moved on.
Then the silence began to break.
Former children spoke. Historians investigated. Museums documented the experiences. Politicians began acknowledging what had happened. Switzerland eventually created a legal framework for compensation. Under the Federal Act on Compulsory Social Measures and Placements before 1981, qualifying victims can receive a CHF 25,000 solidarity contribution recognising the injustice they suffered.
The fact that Switzerland is still officially addressing this history today is itself revealing. The Federal Act on Compulsory Social Measures and Placements prior to 1981 came into force in 2017, establishing a framework for acknowledgement, research, archives, support and a CHF 25,000 solidarity contribution per victim. The Federal Office of Justice continues to administer that framework in 2026. The country did not merely discover an embarrassing historical anecdote. It had to confront the consequences of a system that had once been considered legitimate enough to operate.
But history becomes interesting only when it refuses to stay in the past.
It would be easy to end here. We could tell the story of the Verdingkinder, condemn what happened, acknowledge the victims and congratulate modern society for having moved beyond such practices.
But that would be the comfortable ending. And perhaps the wrong one.
That question did not disappear when child labour laws were passed. It travelled. From the farm to the factory. From the factory to the global supply chain. From the visible workplace to the subcontractor. From the employee to the temporary worker. From the factory floor to the digital platform.
And now, increasingly, from the human worker to the algorithmically managed worker.
The worker has not necessarily disappeared. But the worker has become harder to see.
And that may be the most important transformation of all.
Because the nineteenth-century farmer could see the child working in his field. The factory owner could see the child beside the machine. The modern consumer may never see the human being whose labour makes a digital product possible.
And the coming age of artificial intelligence could take that invisibility to an entirely new level.
The machine has arrived, and it has brought with it a promise that sounds almost entirely different from the world of the Verdingkinder. Artificial intelligence is supposed to make work easier, faster and more productive. It can draft a report in seconds, analyse millions of documents, generate software, translate languages, identify patterns and increasingly perform tasks that once required highly trained professionals. The story we tell ourselves is that technology is liberating human beings from repetitive labour.
But there is another story underneath the spectacular one. Someone still has to train the machine. Someone has to label data, evaluate outputs, identify errors, moderate material, verify information and decide whether an AI system has produced something useful or something dangerously wrong. The machine may appear autonomous. The labour behind it often is not.
This is where the Swiss story becomes unexpectedly relevant. The point is not that a data worker in 2026 is equivalent to a Swiss child labourer a century ago. That comparison would be both historically wrong and intellectually lazy. The people are different. The technology is different. The legal environment is different. The economic circumstances are different. What deserves comparison is something deeper: the relationship between vulnerability, bargaining power and the price of labour.
The nineteenth-century worker could be exploited because he or she was poor, dependent and difficult to replace through anything other than another vulnerable worker. The twenty-first-century worker can face a different kind of pressure. A platform can divide work into tiny tasks. An algorithm can measure performance continuously. A company can outsource employment through layers of contractors. A worker can be located thousands of kilometres from the company whose products depend upon the work. The person remains essential to the process while becoming increasingly invisible within it.
That invisibility matters. The industrial revolution put workers behind factory walls. The globalisation revolution put them behind supply chains. The platform economy put them behind applications. Artificial intelligence may put them behind algorithms. The result is not necessarily less human labour. It can be less visible human labour.
The distinction is becoming increasingly important because the evidence does not support the simplistic claim that AI is about to eliminate most human employment. The International Labour Organization's latest global index estimates that around one in four workers are in occupations with some exposure to generative AI, but concludes that transformation is more likely than outright replacement for most jobs. Its June 2026 review finds that productivity gains are real but uneven, while the more immediate risks include inequality, weaker job quality, declining worker autonomy and disruption for particular groups, including younger workers.
That makes the real question more interesting than “Will AI take our jobs?” The real question is what kind of work will remain, who will perform it, how much bargaining power they will have, and who will capture the productivity gains created by the machines.
That is the question the Verdingkinder history helps us ask.
A technological revolution does not automatically create a fairer labour market. Neither does economic growth. Neither does higher productivity. The gains from technological change have always depended on institutions, bargaining power, regulation and the distribution of economic ownership. The spinning machine did not decide what factory workers should earn. The factory owner did. The law did. The labour movement did. Society did. AI will not decide what happens to workers either. We will.
That is why the debate about artificial intelligence cannot be reduced to technical capability. A machine may be able to perform a task. That does not tell us whether the task should be automated, whether the worker should be displaced, whether the remaining work should become more valuable, whether wages should rise, whether working hours should fall or whether the productivity gains should flow almost entirely to the owners of the technology.
Those are political and economic choices.
And history gives us a warning about what happens when those choices are made without giving the vulnerable worker a meaningful voice.
Consider the difference between a worker who can negotiate and a worker who cannot. A worker with bargaining power can say: “This wage is too low.” A worker with legal protection can say: “These conditions are unacceptable.” A worker with an alternative employer can say: “I am leaving.” A worker with collective representation can say: “We will not accept this.” But what happens when the technology itself weakens those alternatives?
That is the question algorithmic management raises.
A worker whose performance is continuously measured by software may be told how quickly to complete a task, how many tasks to handle, when to log in, when to log out and how to respond to customers. The manager becomes less visible because the algorithm becomes the manager. The worker may technically remain employed, yet the relationship can become increasingly mechanical.
The International Labour Organization has already identified algorithmic management and the role of the “data labourers” who underpin AI systems as important dimensions of the changing world of work. Its research also warns that AI adoption can affect job quality, worker autonomy and working conditions depending on how the technology is introduced.
This is where the old vocabulary begins to return in new forms: efficiency, flexibility, productivity, optimisation, cost reduction, scalability.
These are not bad words. In many circumstances they describe genuine economic progress. But history teaches us to ask a second question whenever those words appear: Efficiency for whom?
If an algorithm allows one worker to produce twice as much, that can be wonderful. Perhaps the worker finishes earlier and earns more. Perhaps the company becomes more competitive and creates new jobs. Perhaps customers receive better products at lower prices.
But there is another possibility. The company keeps the productivity gain. The worker is expected to produce twice as much. The number of workers is reduced. The remaining workers become more closely monitored. And the productivity revolution becomes, from the worker's perspective, a demand to work harder with fewer people.
The technology has increased productivity. But it has not necessarily increased power. That distinction could define the social history of AI.
The great promise of artificial intelligence is that it can increase the amount of economic value produced by each hour of human effort. If that happens on a sufficiently large scale, humanity faces an extraordinary opportunity. We could produce more with less labour and use the resulting wealth to shorten working hours, increase incomes, expand education, strengthen social protection and give people more freedom. The crucial word is could. There is nothing automatic about that outcome. The alternative is equally possible: a small number of firms capture a disproportionate share of the gains while workers compete for the shrinking pool of tasks that machines cannot yet perform cheaply. Then the problem is no longer simply unemployment. It is bargaining power. A worker does not need to lose a job completely to become economically weaker; the possibility that a machine could perform the task may itself reduce the worker's negotiating position. That is why the most important question of the AI economy may not be how many jobs disappear. It may be how much the threat of disappearance changes the value of the jobs that remain.
The Verdingkinder offer a grim historical illustration of the same underlying principle from a radically different era. Their vulnerability was part of what made their labour exploitable. They could not negotiate on equal terms. They could not easily leave. They depended upon institutions and adults who possessed vastly greater power.
Modern workers are not children being auctioned at a farm. But a worker who cannot afford to refuse a contract, cannot survive a period without income, cannot challenge an algorithmic decision and cannot negotiate with the platform controlling access to work is also operating from a position of vulnerability.
The forms are different. The underlying economic question is familiar. How much freedom does a worker really possess when saying no carries a price they cannot afford?
That question becomes even more important as AI spreads across borders.
The benefits of AI will not be distributed evenly because the technology itself is not entering equal societies. A joint ILO–World Bank study covering 135 countries warns that developing economies may experience disruption before seeing comparable productivity gains because of digital gaps and differences in the tasks workers perform. At the same time, the World Bank's 2026 World Development Report argues that AI could become a major productivity opportunity for developing economies if they close gaps in electricity, connectivity, skills and institutional capacity. The future is therefore not predetermined: the same technology can widen inequality or accelerate development depending on whether societies build the conditions that allow workers to benefit.
That creates the possibility of a new global division of labour. At the top will be the people who own the models, computing infrastructure, data, intellectual property and capital. Below them will be highly skilled workers who use AI to multiply their productivity. Below them may be millions of workers performing the tasks that AI cannot yet fully automate. And beneath even those workers may be the invisible labour required to keep the systems functioning: data preparation, content moderation, evaluation, verification and other forms of digital work.
The paradox is striking. The more artificial intelligence appears to become, the more important it may become to ask where the human labour has gone.
A photograph of a nineteenth-century factory could show us the child standing beside the machine.
A modern AI product may show us nothing.
There may be no photograph of the worker who helped train the system. No photograph of the person who reviewed thousands of outputs. No photograph of the moderator who filtered disturbing material. No photograph of the contractor working through a platform in another country.
The labour disappears from the consumer's field of vision. And when labour disappears from view, responsibility can disappear with it. This is one of the great dangers of the AI economy: not that machines will necessarily become cruel, but that humans may become increasingly comfortable with economic arrangements whose human costs they no longer have to see.
That is why the Swiss story deserves to be remembered.
The Verdingkinder system was not sustained because every person involved was uniquely monstrous. It survived because an institutional arrangement could distribute responsibility so widely that the exploitation became easier to normalise. The farmer could say the authorities had approved the placement. The authorities could say the child was being provided for. Society could say the child was being taught discipline. And the economic system could quietly benefit from the labour.
Nobody needed to believe that they were participating in a system of exploitation. The system simply had to make exploitation useful.
That is a lesson for the age of AI.
When an algorithm reduces costs, someone benefits. When a platform transfers risk to workers, someone benefits. When a company replaces ten tasks with one system, someone captures the resulting value. When AI increases output without increasing wages, someone receives the productivity dividend. The critical question is therefore not whether technology is good or bad. It is: Who has the power to decide what technology does to the worker—and who gets the productivity dividend?
That question takes us back to the child.
The Verdingkind could not negotiate the terms of the arrangement into which adults had placed them. The modern worker must be able to.
That means the future of AI cannot be built around productivity alone. It has to be built around human bargaining power. Workers need a voice in how AI is introduced. They need protection against arbitrary algorithmic decisions. They need opportunities to acquire new skills. They need social protection during transitions. They need the ability to challenge systems that determine their livelihoods. And where AI produces extraordinary productivity gains, societies need to confront the question of how those gains are shared.
That is the opposite of the logic that once governed the Verdingkinder.
The child had almost no say. The future worker must have one.
Because there is a profound difference between technology replacing human effort and technology reducing the value of human beings.
Artificial intelligence could become the greatest productivity technology since the Industrial Revolution. It could remove dangerous work, reduce drudgery, expand access to knowledge and allow humanity to produce extraordinary amounts of wealth with less human effort. The humane response to that possibility is not to defend every existing task merely because it is human. It is to ensure that when machines make some work cheaper or unnecessary, human beings become freer rather than simply more disposable. If the gains are captured narrowly while the risks are pushed downward, AI could instead produce a labour market in which millions remain economically necessary but increasingly powerless. That would be a very different kind of technological revolution.
The machines would be intelligent. The system would not necessarily be.
And perhaps that is the real lesson buried inside Switzerland's forgotten children.
A society should not judge its economic progress only by what its machines can produce. It should ask what happens to the people whose labour the machines make cheaper, whose skills the machines make less valuable, and whose bargaining power the machines may quietly erode.
Because the most dangerous form of exploitation is not always the one that looks brutal.
Sometimes it looks like efficiency. Sometimes it looks like innovation. Sometimes it looks like flexibility. Sometimes it looks like progress.
And sometimes it arrives disguised as a machine that promises to make human labour obsolete.
The nineteenth-century child was visible in the field. The factory child was visible beside the machine. The modern digital worker may be visible only in a database. The AI worker may be visible only as a stream of labelled data, a moderation decision, an evaluation score or a completed task.
The human being disappears. The economic value remains.
The machines are new. The algorithms are new. The vocabulary is new. But the question is ancient: Who benefits when human vulnerability becomes economically useful?
History does not repeat itself.
It changes vocabulary.
And the challenge of the AI age is to make sure that the vocabulary of innovation never becomes another respectable language for exploitation—and that the productivity of machines becomes a source of human freedom rather than a new measure of human disposability.
Explain It Clearly examines complex developments in geopolitics, economics, technology, education and society through clear, accessible and independent analysis.
Comments
Post a Comment