What if we’re reading the AI revolution all wrong?
Revolution, after all, has meanings beyond impactful technological advancements. Rather than explosions in productivity and creativity, revolution might connote explosions of the car-bomb and Molotov cocktail variety. In a particularly dark vision of the future, AI causes so much civil, political, economic and cultural disruption that it spills over into unrest.
That scenario, to us, has sounded like a pretty steep slippery slope until this week, when Microsoft co-founder Bill Gates (of all people) decided to warn of the disruptive potential of AI.
Gates has spent much of the artificial-intelligence boom arguing that AI could become one of humanity’s most consequential technologies for improving health, education, agriculture and productivity. That makes the warning he issued this summer unusually significant. Gates now describes the transition into an AI-dominated economy as potentially one of the most turbulent periods in human history and says that, if somebody could devise a credible global mechanism for slowing AI development, he would probably support it.
“If someone had a credible plan for slowing down AI advances globally, I would likely support it,” Gates wrote, while conceding that geopolitical and economic competition makes such an agreement unlikely. More strikingly, Gates warned that governments need time to prepare for “social, political, and economic upheaval” as artificial intelligence begins substituting for human cognition across industries.
That is a remarkable statement from the Microsoft co-founder, but it increasingly sounds less like an outlier than an expression of a political problem that has been developing throughout the summer. Anxiety about artificial intelligence has increasingly escaped the confines of AI-safety conferences, technology publications and Washington policy debates. AI has become something people can oppose politically. Data centers are being protested. Workers are demanding contractual protections from automation. Candidates are discovering that opposition to AI infrastructure can attract voters. Environmentalists, populist conservatives, organized labor, progressives, artists and young workers are arriving at overlapping complaints about the technology from very different ideological starting points.
None of this means an “AI revolt” is imminent. There is little evidence that artificial intelligence itself is presently causing widespread civil disorder. But several mechanisms through which technological change historically produces unrest — employment displacement, inequality, geographically concentrated economic disruption, perceived loss of political control and rapidly spreading misinformation — are beginning to converge around AI.
For financial services leaders, this distinction matters. The important risk may not be machines suddenly becoming hostile to humanity. It may be humans becoming hostile to an economic order they believe is being reorganized around machines.
From AI Anxiety to AI Backlash
The simplest evidence for an emerging backlash is public opinion. Pew Research Center reported Aug. 18 that 52% of American adults now say they are more concerned than excited about increased AI use, compared with 37% in 2021. Only 9% are more excited than concerned. More revealingly, 71% expect AI to result in fewer U.S. jobs over the next 20 years, compared with 64% in 2024. Among Americans under 30, 73% expect fewer jobs and 55% now describe themselves as more concerned than excited about AI.
That younger cohort deserves particular attention. Earlier this year, commencement speakers at several universities were booed when they enthusiastically discussed AI, including former Google CEO Eric Schmidt at the University of Arizona. Those episodes were small cultural events rather than political demonstrations, but they suggested something important: the generation assumed to be most naturally comfortable with artificial intelligence may instead associate it with diminished employment prospects, weakened creativity and a future designed by older technology executives.
Economic evidence remains much less dramatic than public expectations. Goldman Sachs research reported in August found “visible but narrow” labor-market effects from AI so far, with AI-sensitive industries showing weaker hiring and entry-level workers appearing particularly vulnerable. The difference between current economic evidence and public expectations is important. Political movements frequently form around anticipated losses, not merely losses that have already occurred.
The intellectual backlash is also broadening. Writing in Jacobin in July, economist Branko Milanovic returned to Marx to examine what happens when increasingly capital-intensive AI production substitutes machines for labor. The larger debate extends well beyond Marxism: economists are asking who owns AI-generated capital, who receives the productivity gains and how governments finance themselves if labor receives a declining share of national income. A July NBER paper similarly explored fiscal responses to scenarios involving greater productivity, inequality, displacement and an increasing capital share of income.
That debate has an obvious political destination. If voters eventually perceive AI as something that eliminates their jobs while enriching technology companies and investors, AI ceases to be primarily a technology issue and becomes a distributional one.
The Data Center Becomes AI’s Political Symbol
Artificial intelligence is intangible. Data centers are not.
They occupy hundreds of acres, consume electricity and water, require transmission infrastructure, receive tax incentives and can produce noise and construction disruption. That makes them an ideal physical target for dissatisfaction with a technological transformation that otherwise happens invisibly inside models and software.
The environmental questions are substantial. Our World in Data estimates that data centers currently consume around 1.5% of global electricity, with demand heavily concentrated geographically. Research highlighted by North Carolina State University projected that expanding data-center and cryptocurrency demand could increase average U.S. electricity costs by 6% to 29% by 2030 under modeled scenarios, with much larger increases possible in some regions.
The economic case is more complicated than either advocates or opponents sometimes acknowledge. Georgia Tech researchers found measurable increases in employment, wages, businesses and household income following data-center openings. But benefits were disproportionately captured by metropolitan areas. Rural counties often received fewer spillovers because permanent employment at facilities can be relatively small and specialized contractors may come from elsewhere. The same study found electricity prices rising about 5% where effects could be cleanly measured.
That combination — enormous investment, modest permanent employment, resource consumption and potentially higher household costs — is politically combustible.
By July 18, coordinated demonstrations against data-center expansion reportedly encompassed 142 protests across 42 states, illustrating how quickly scattered local disputes could become a national movement. By late August, data-center controversies were affecting midterm-election politics, while Republicans and Democrats alike were confronting voters angry about electricity rates, water consumption, land use, noise and corporate subsidies.
The backlash is cultural as well as ecological and economic. Communities frequently frame the fight as one of local control against remote corporations. Artists, writers and cultural institutions object to human creativity being absorbed into training datasets. Creative Commons warned Aug. 19 that commercial AI practices involving cultural heritage can conflict with norms of attribution, provenance, reciprocity and openness, and noted that some institutions have responded to AI scraping by restricting access to collections.
The emerging complaint is therefore broader than “AI uses too much electricity.” It is that corporations are appropriating electricity, water, land, data, creative work and human knowledge while the public receives an uncertain share of the gains.
Anti-AI Populism Comes From Both Directions
This creates unusually fertile territory for populist politics because opposition does not fit cleanly onto America’s left-right axis.
On the left, Sen. Bernie Sanders has gone furthest. He proposed giving the public substantial ownership of large AI companies through a sovereign wealth fund, has joined Rep. Alexandria Ocasio-Cortez in backing a data-center moratorium and on Aug. 10 called on Anthropic, Meta and OpenAI to pause advanced AI development. The underlying progressive argument combines traditional concerns about concentrated corporate ownership, workers’ bargaining power, environmental justice and economic inequality with newer AI-safety concerns.
Sanders’ sovereign-wealth proposal is particularly interesting because it converts AI resentment into a question familiar to financial professionals: ownership. If machines increasingly generate the economy’s output, should households receive more income from capital? The broader debate over distributing AI wealth — through taxation, sovereign ownership, dividends, stronger labor protections or other mechanisms — is moving from economic speculation toward practical politics.
But Republican Sen. Josh Hawley has developed his own version of anti-AI populism, emphasizing family protection, surveillance, electricity costs, employment and Big Tech power. Hawley has argued that workers need rights over workplace AI, backed restrictions designed to prevent autonomous trucks from replacing drivers and sponsored bipartisan legislation intended to prevent data centers from shifting electricity costs onto households.
This does not mean the parties have converged. President Donald Trump remains strongly supportive of AI infrastructure and data-center construction as an economic and strategic priority, particularly in competition with China. Labor itself is divided: construction unions can benefit enormously from data-center building, creating political tension with white-collar workers worried about AI displacement and communities opposing the facilities.
The better conclusion is that there is currently no coherent “left” or “right” response to AI.
The left’s emerging opposition emphasizes labor displacement, inequality, climate and concentrated corporate ownership. The right’s emphasizes household costs, surveillance, children, local sovereignty and hostility toward Big Tech. Yet those arguments overlap around a populist narrative: powerful corporations are transforming society without meaningful democratic consent.
That overlap explains why data centers have produced strange coalitions of environmentalists, rural landowners, progressives and conservatives. AI could become one of the rare political questions on which ideological opponents agree about the problem while disagreeing profoundly about the solution.
The Misinformation Accelerator
There is another route from AI to unrest that requires no mass unemployment whatsoever: information.
Generative AI dramatically reduces the cost of producing persuasive false text, photographs, audio and video. The danger is not simply that everyone believes deepfakes. Research increasingly suggests something subtler and potentially more corrosive.
A 2026 systematic review of political deepfake research found that deepfakes can alter perceptions, create confusion and reduce trust in both news and government, although they are not necessarily more persuasive than conventional misinformation. Another systematic review described synthetic media as creating a “skepticism tax”: the damage occurs partly because citizens become less certain that authentic evidence can be trusted.
Experiments in the United States and Netherlands found political deepfakes could reduce support for targeted politicians even though viewers rated the fabricated videos as less credible than authentic ones, with some delegitimizing effects persisting despite corrections. Research published July 14 involving 7,000 respondents in Kenya further demonstrated how surrounding social commentary can affect perceptions of whether fabricated political video is authentic.
This is where misinformation intersects directly with unrest. During an election dispute, war, terrorist attack, police confrontation or protest, AI-generated imagery no longer has to convince an entire population. It merely has to reinforce the beliefs of an already angry group, manufacture apparent evidence of misconduct, overwhelm verification systems or create enough uncertainty that people choose whichever version of events fits their political identity.
Research publicized in August adds another complication: large language models can act as “ideological chameleons,” adjusting their framing of contentious political issues toward users’ disclosed ideological positions. Researchers warned that such behavior could contribute to polarization.
The Brennan Center for Justice demonstrated the duality in August tests of major chatbots. Many systems challenged election conspiracy theories and refused requests to facilitate deceptive election content, suggesting AI can be a misinformation defense. Yet researchers also showed that generative systems can dramatically reduce the technical barriers to producing sophisticated deceptive media.
The danger is therefore not an AI system independently deciding to start a riot. It is AI becoming an accelerant inside an already combustible political environment.
From Inequality to Instability
Perhaps the most provocative recent research comes from economists thinking explicitly about automation and political stability.
A June NBER working paper by Daron Acemoglu, A. Arda Gitmez and Mehdi Shadmehr modeled automation in a political system where workers can revolt. In their theoretical framework, automation increases capital’s share of national income and greater inequality raises the probability of revolt, forcing governments to choose among regulating automation, redistributing income and repressing political action.
That is a model, not a prediction that Americans will soon take to the barricades because of ChatGPT. But its logic identifies the critical variable: distribution.
Other recent scholarship has similarly linked AI-driven inequality, fiscal stress and institutional fragility to potential instability, particularly in already vulnerable societies. The World Economic Forum’s 2026 Global Risks Report describes a possible feedback loop in which labor displacement increases inequality, inequality intensifies polarization, misinformation erodes institutional trust and governments respond to instability with greater surveillance. It specifically warns that prolonged unemployment among educated young people could become socially destabilizing.
Gates arrives at much the same problem from a very different ideological direction. Employment, he argues, provides not only income but dignity and social connection. If AI rapidly removes entry- and mid-level jobs while economic gains accrue disproportionately to a small group, governments could confront structural rather than cyclical unemployment. His proposed responses include stronger safety nets, reserving certain occupations for humans and taxing AI tokens and robots partly to discourage excessively rapid substitution of capital for labor.
Most importantly, Gates explicitly warns that governments should act before unemployment rises, communities suffer and public trust erodes. Without solutions, he says, groups will eventually demand that AI development and deployment be stopped altogether.
The Coming Politics of AI
Will anti-AI sentiment grow?
Probably, unless ordinary people begin experiencing AI primarily as something that improves their economic circumstances rather than threatens them. The current polling trajectory is unfavorable to the industry. Concern is rising, particularly among young adults, while expectations of employment destruction are becoming mainstream. Data-center opposition has already crossed from NIMBY-style zoning disputes into electoral politics.
But political unrest is not inevitable, and neither is mass unemployment. AI could generate productivity, new businesses, better healthcare, cheaper professional services, scientific breakthroughs and entirely new categories of employment. The World Bank argued this month that AI could substantially accelerate development in poorer countries if governments address gaps in electricity, connectivity, skills and institutional capacity.
The crucial question is whether those benefits become visible quickly enough, and broadly enough, to establish political legitimacy.
In the near term, AI-related civic action is much more likely to resemble what is already happening: demonstrations against data centers, zoning fights, lawsuits, moratorium campaigns, labor bargaining, electoral challenges, calls for taxes and restrictions, and populist politicians attacking technology companies. More than 15 political candidates have already signed an “AI Pact” promising action on AI safety and data centers.
More serious unrest becomes plausible if several conditions arrive simultaneously: visible AI-driven layoffs, a recession or weak labor market, rising electricity costs, conspicuous technology-sector wealth and a political or geopolitical crisis saturated with synthetic misinformation. Under those circumstances, AI could provide both the grievance and the communications technology through which that grievance is organized and amplified.
That possibility should matter enormously to financial institutions. Banks, asset managers, insurers and wealth managers sit close to virtually every fault line: capital ownership, employment automation, market concentration, household wealth, cybersecurity, data-center financing and the distribution of investment gains. They may therefore discover that “AI risk” encompasses considerably more than hallucinations, model governance and cybersecurity.
The larger lesson of the past seven weeks is that the social license for AI cannot be taken for granted. The technology industry has largely framed the AI race in terms of capabilities, investment, productivity and national competition. The public is increasingly asking a different set of questions: Who gets the jobs? Who gets the wealth? Who pays the electricity bill? Who controls the data? Who decides what gets automated? And who has a voice when the answers affect an entire community?
Those are not engineering questions. They are political questions.
And if governments, companies and financial institutions cannot produce convincing answers, artificial intelligence may eventually become more than another issue in an already polarized political system. It could become a powerful organizing symbol for dissatisfaction with the economic system itself.
Researched by DWN Staff
