Could Artificial General Intelligence Adoption Start a Civil War in America?
With indicators already pointing to a volatile level of political stress in the United States, a large-scale displacement of educated elites by artificial general intelligence (AGI) could spark a convulsive breakdown of the state.
Technological innovation has been a major driver of productivity and economic growth since the start of the industrial age. However, gains have almost invariably come at the expense of workers with lower skill and education levels. As but one example, automation from 1990-2007 resulted in the loss of an estimated 400,000 jobs in the United States, primarily among non-college educated workers.
Educated elites often avoided large-scale job displacement from technological innovation in part by possessing difficult to automate skills. But their vulnerability could increase if the most visionary ideals of artificial intelligence (AI) come to pass. Already, analysts warn that white collar jobs could be threatened by generative AI technologies, which have shown an impressive ability to carry out tasks involving images, video, and text. A Pew study concluded that one in five American jobs have a “high exposure” to artificial intelligence, with high earners and the college educated most exposed. The situation for educated elites could worsen significantly. Some experts have predicted that “artificial general intelligence” (AGI) may soon be available. Although prospects for and definitions of this potential technology remain hotly debated, AGI may be generally understood as a versatile, reliable general intelligence that is comparable or superior to human intelligence.[1] A report by Goldman Sachs assessed that AGI capable of carrying out work “indistinguishable” from human output could wipe out a fourth of current jobs and 300 million jobs worldwide.
The bitter irony that educated elites could be rendered obsolete by the advanced technologies that they once hailed may well inspire a sense of just desserts, especially among the workers who have long borne the heavy cost of “technological progress.” But however much one may be tempted to relish the misfortunes of the well-to-do, the political repercussions of such an outcome are potentially catastrophic for the entire country.
Political scientist Peter Turchin has provided the most thorough explanation of why a collapse in elite employment could be disastrous. Turchin’s research, which builds on the work of George Mason University political scientist Jack Goldstone, shows that the combination of popular immiseration and elite overproduction has often resulted in a breakdown of the state. Popular immiseration is when non-elites (bottom 90% of income earners) experience stagnating or declining living standards. Immiseration raises discontent, lowers trust in the government, and elevates the potential for anti-system political mobilization. Elite overproduction is when the supply of positions that support an elite lifestyle is greatly outnumbered by aspirants to those positions. Because elites have access to resources and networks in key sectors of state power, including the military, government, and economy, their opposition to the state can pose a serious threat. Frustrated elites can become “counter elites” who exploit popular grievances to lead rebellions or revolutions against authorities. Turchin and Goldstone grimly warn that in past societies facing such predicaments, about 75% of the cases resulted in some sort of political collapse, such as civil war or revolution.
A review of the situation in the United States through this lens provides ample cause for alarm. Popular immiseration has become a dispiriting feature of American society. Income and wealth inequality has worsened as the non-elite share of national income and wealth has shrunk. Indeed, the fortunes of the working class have largely stagnated or worsened since the 1970s. The elimination of good paying industrial jobs via automation proved a major contributor to their declining fortunes. Symptoms of popular immiseration include the increasing gap in life expectancy and health outcomes between the working class and college educated elites. While college educated people have become healthier and wealthier, Americans without college degrees are dying from pain and despair. Not surprisingly, popular trust in the government and confidence in the future has plummeted, steadily declining since the 1970s and rarely reaching above 20% since 2008.
Non-elites have responded to their worsening situation by repeatedly expressing their dissatisfaction through their votes and via mass protests. Waves of popular unrest have roiled modern societies since the turn of the century, and some political scientists warn that the severity of anti-government demonstrations will likely intensify in the future as global economic activity slows and the international security situation deteriorates. However, non-elites have generally lacked the organization and resources to reverse their situation and restore their fortunes. Elites, who have usually favored and benefited from automation trends, are generally too entrenched in the centers of state power to be easily overcome.
The failure to address the grievances of the working classes leaves a persistent peril, however. As trust has evaporated and discontent grown, countries have become less stable. As conditions for non-elites become intolerable, more seek to join the already swollen ranks of elite aspirants. Even so, elites can generally manage popular discontent so long as they work together. The danger increases, however, once elites are no longer unified. Although elites routinely divide themselves into factions and groups and disagree on many issues, the most profound and bitter divisions emerge over access to resources to support elite lifestyles. Although there are always more elites and elite aspirants than there are positions to support them, “elite overproduction” occurs when the supply of elites vastly outnumbers available positions. Past incidents of civil war and political revolution have underscored the critical role that divided elites play in the onset of mass political violence.
Signs that elite aspirants already exceed available positions can be seen in the decline in employment opportunities for individuals with college and advanced degrees in all countries. As an example, the underemployment rate for American college graduates has exceeded 50% even ten years after graduation, with many unable to pay off school debts. The increasingly ruthless struggle for admissions into elite schools, and the collapse of the humanities departments in favor of business and science, technology, engineering, and math (STEM) programs that promise better job prospects also reflect the same problem of overproduction of elites. Educated European youth have faced elevated unemployment rates for years, reaching 15% in 2020. Although jobless rates have since subsided, many young Europeans face high prices with precarious jobs. Across the European Union, nearly a third live at home to save money. In some countries, such as Croatia, the number is closer to 65%. The problem of surplus educated youths has intensified in China as well, especially as its economy cools. Unemployment levels for educated urban youth has surpassed 20% since 2020. In 2011, the death of a Tunisian fruit seller kicked off a wave of protests among young people. The protests reverberated around the world, sparking a historically precedented level of unrest around the world, especially among educated youths in Europe, the Middle East, North Africa, Latin America, and the United States.
The problem of elite overproduction and popular immiseration occur along signs of state breakdown around the world. Globally, domestic instability and violence have been on the rise. Liberal and democratic freedoms and institutions have deteriorated steadily for 18 straight years. Although conflict between states has increased in the past three years, violence between domestic groups remains far more common. Fighting within states tends to center around issues of the breakdown in the rule of law, resource scarcity, and illicit economic gains. The number of conflicts within a state has since increased steadily since the end of the Cold War, peaking in 2016 and subsequently subsiding slightly. In 2023, 75 countries experienced intrastate conflict, resulting in 12,000 deaths.
AGI’s Possible Impact on Political Stability
The widespread adoption of transformative AI could drive countries already grappling with fragile political situations into crisis or collapse. As a general purpose technology that could execute intellectual, analytical, and cognitive tasks better than trained humans, AI could render large numbers of educated elites redundant, possibly even before the technology crosses the unknown, possibly unknowable threshold to AGI. Professionals such as lawyers, analysts, computer programmers, financial services, and others could be replaced by AGI powered computers and robots. Although the advent of AGI could create some new jobs, the advent of information technologies in the 1990s suggests the number of new, good paying job may be far smaller than the large number of jobs lost through automation. With much of the populace already disaffected, elites frustrated by the prospect of downward mobility could have a strong incentive to oppose leaders who backed AGI adoption. Counter-elites could find ready audiences among a disaffected populace that has seen little improvement in its situation. Efforts by the state to suppress insurrections led by counter elites could trigger civil war or revolution.
Liberal democratic states such as the United States could be especially vulnerable to AGI-driven instability. These governments are less likely resort to the brutal and cruel measures frequently employed by autocracies to control numbers of elite aspirants, such repression, purges, and arbitrary violence. This still leaves unresolved the problem of an oversupply of elites. The experience of democracies in the developing world, where elite aspirants have long outnumbered resources to support them, suggest one disturbing possibility. In these countries, rival factions sometimes weaponize political institutions to suppress their rivals and capture the spoils of the state. Liberal democracies may similarly experience a growing temptation to adopt illiberal practices to marginalize and suppress rival elites. The fraying of democratic norms, increasing resort to hardball politics, the spread of political violence, and general decline in popular support for democratic institutions in the wealthiest and most stable democracies suggest that these trends may already be well underway.
Conclusion
Options to mitigate the danger of political upheaval owing to AGI-driven job losses among elites are few and difficult to implement. The most effective measures would address the core drivers of popular immiseration and the overproduction of elites. But past efforts to adopt ameliorative measures to ease such political pressures have rarely succeeded.
If provided enough basic goods and services, popular immiseration could be reversed or at least eased. This could have the salutary effect of reducing potential constituencies for counter-elites. However, policies to ensure access to a basic level of basic goods and services such as food, housing, basic education, and health care for most of the population would be extremely expensive. Efforts to fund such measures through taxation of the wealthy have generally failed, as elites have historically proven effective at stopping such transfers owing to their dominance of state power. Indeed, Stanford University professor Walter Scheidel has documented how high levels of inequality have historically only been overcome through violent cataclysms such as lethal pandemics, wars, communist revolution and societal collapse.
There are also few options to resolve the problem of elite overproduction. One option could be to restrict access to elite positions. China, fearing the potential destabilizing effects of unemployed educated youth, has adopted measures to restrict access to college and even high school for the country’s youths and to instead compel more workers to accept jobs in trades and the lower paying service sectors. These measures have encountered fierce resistance by many parents angered by state-directed efforts to foreclose possibilities for their children. Free societies are even less likely to tolerate such restrictive measures.
Buying elite support through state subsidy is another option. Some countries, such as China, rely on state employment as a way to improve political stability. In the United States, an oversupply of college graduates has led many job postings to require post-secondary education that previously did not require it, raising charges of “degree inflation.” However, state subsidies to an excessively large pool of elites could also prove extremely expensive and it could incentivize more people to become elite aspirants, further aggravating the problem of elite overproduction. Moreover, this option carries the risk that control of the state will become viewed as a prize worth fighting for as has occurred in some developing countries.
Analysts have long discussed the issue of AI “alignment” with the values of the “human race.” But such abstract and philosophical discussions, however important, overlook the far more urgent and proximate question of a technology’s “alignment” with the interests of contending groups of people in any given society. Technological innovation has almost always had an “alignment” problem in this sense: some groups have invariably gained while others lost. The advent of AGI will differ only in the identities of the potential gainers and losers. The promise of super productivity through widespread adoption of AGI could well be a boon to mankind. But without a careful consideration of the potential costs, the potential for disaster suggests more thought is urgently needed to better understand the technology’s attendant risks, as well as its promised opportunities.
[1] See, for example, Benjamin Boudreaux, Taking Artificial General Intelligence Seriously, Not Literally, Santa Monica, CA: RAND Corporation, 2025 (Forthcoming). Boudreaux emphasizes the three dimensions of capabilities, deployment, and emergence when thinking about AGI.
