The launch of ChatGPT in 2022 unleashed the rise of artificial intelligence and triggered a flood of warnings from industry leaders about a possible labor apocalypse. Although they have reasons to highlight the disruptive effect of their products, and despite employment in developed countries being near historic highs, the pessimistic message has taken hold. Seven out of ten Americans believe that artificial intelligence will make it harder to find a job; nearly a third fear for their own employment. The lack of opportunities for university graduates — especially computer programmers — increases concern.
Read more The unanswered questions left by the trial against the Pujol family
The past offers some comfort to the worried. Labor markets are constantly changing. Today’s offices would be unrecognizable to a worker from 50 years ago. Never in modern history has technological progress hampered overall labor demand. Economic historians now downplay the magnitude of the so-called “Engels pause,” the period of the Industrial Revolution when working-class wages grew more slowly than the overall economy.
However, history is not always a good guide to anticipate the future, as the Industrial Revolution itself demonstrated. The most advanced artificial intelligence models are impressive. They can tackle programming tasks much more complex than was predicted a year ago. The number of AI agents has skyrocketed. Corporate spending on artificial intelligence has increased significantly. Anthropic’s annualized recurring revenue, one of the leading companies in this field, could reach $50 billion by the end of June. There is still no evidence in labor market data that AI is destroying many jobs. However, given how quickly it is advancing, it would be unwise to dismiss fears that it might. Society could be on the brink of a profound resource reallocation and political upheaval.
Over time, humans could, like horses during the automobile era, cease to be profitable
The economists’ prediction that work will remain abundant is less reassuring than it seems, especially in the long term. Although the market may find uses for human labor even as models and robots become increasingly capable, the quality of those jobs and the wages they offer are not guaranteed. Goldman Sachs, a bank, forecasts that data centers will account for 8.5% of peak energy demand in the United States in 2027, up from 4.1% in 2025. As AI companies raise the price of land and energy, the money people earn will go further less. Over time, humans could, like horses during the automobile era, cease to be profitable. Income could end up going mostly or entirely to those who own capital, who would then spend it on goods produced by artificial intelligence and robots, using natural resources they control.
This dystopian possibility underlies Silicon Valley’s warnings about the need for state intervention and even the implementation of a universal basic income. That remains something distant, if it ever becomes reality. However, governments may need to act sooner, because a cataclysm is not necessary to fuel popular anger. It is estimated that about two million Americans lost their jobs between 1999 and 2011 due to China’s entry into the global trading system. That figure is no worse than the number of layoffs in a typical month in the turbulent U.S. labor market. However, the “China shock” helped bring Trump to the White House and led to the imposition of the highest tariffs since the 1930s.
Office workers threatened by artificial intelligence have more political and social weight than factory workers harmed by Chinese competition. Even a small number of layoffs could provoke a backlash against technology; fierce opposition to new data centers is a sign of what could happen. A severe disruption of many people’s security and status could lead to widespread unrest, and even revolution.
What should governments do? One proposal is to slow down change. China, for example, has urged its companies to adopt artificial intelligence but without laying off employees. Prestigious economists worldwide have suggested raising taxes on capital and reducing the tax burden on labor. Some activists, meanwhile, advocate taxing data centers. However, slowing technological progress is not the most sensible path. Humanity is likely to gain enormous benefits from artificial intelligence: not only more wealth but also advances in fighting diseases and solutions to problems like climate change or poverty. If the Luddites had succeeded in preventing the automation of looms in early 19th-century England, the world would be in a much worse situation today.
Read more ERC convenes its national council this Monday to assess the approval of the budgets
A second category of compensatory measures would be preferable. If employment declines, the income workers previously received will likely show up as large profits in AI companies, chip manufacturers, data centers, or other points in the supply chain. Smart tax reforms, such as taxes on corporate profits above a normal return on capital, on land, or on natural resources, could capture these rents. The argument for inheritance taxes to prevent the consolidation of a capital-owning elite now seems even stronger.
At the same time, governments could help workers adapt. Public wage insurance systems, which smooth income loss after job loss, can help workers find better opportunities (and thus eventually become self-financing). Denmark’s active employment policies, where the state helps people search and train for new occupations, have been shown to reduce unemployment durations.
Also read
These ideas would make the economy more efficient and fair regardless of artificial intelligence. Would they be enough to convince voters facing uncertainty and disruption? In a populist era, technocratic reforms are hard to sell. Previous attempts to help workers adapt to trade liberalization failed to prevent the reaction to the “China shock.” In an economy fully dominated by artificial intelligence, humans will need help to survive, not just to adapt.
Hence arises a final set of radical ideas, such as partial nationalization of AI companies. This week, a South Korean presidential adviser suggested a “citizen dividend” from AI companies, which caused a 5% drop in the local stock market before being reversed. In the United States, some politicians whisper about distributing AI company shares to citizens through “Trump accounts.” Economically, there is little difference between a well-designed tax system and state ownership in the private sector. Countries without AI giants will have to rely on taxes rather than appropriating shares of foreign companies. However, the U.S. may find that some degree of public ownership is the best way to make the social benefits of this technology transparent.
It is necessary to address rent concentration from the start, before rentiers’ power becomes excessive. The employment apocalypse has not yet arrived. But if governments wait for conclusive evidence before creating a safety net, it will be too late. It is better to start now.
Read more Does the Govern have «leeway» to improve the working conditions of teachers?
© 2026 The Economist Newspaper Limited. All rights reserved.
