global geopolitics

Why Growing Scepticism Towards Artificial Intelligence Reflects Political Economy Rather Than Fear of Technology

Artificial intelligence has rapidly assumed a position once occupied by electrification, railways, telecommunications and the internet as the defining technology of its generation. Governments describe artificial intelligence as essential national infrastructure. Financial markets increasingly value companies according to their AI strategies rather than existing earnings. Technology executives speak confidently about productivity revolutions, scientific breakthroughs and economic transformation measured across decades rather than business cycles.

Every major technological revolution has been accompanied by promises of shared prosperity. Railways, electrification, mass manufacturing, personal computing and the internet each promised productivity gains that would ultimately improve living standards across society. Artificial intelligence has inherited the same optimistic vocabulary. Much contemporary commentary therefore treats growing public scepticism as irrational resistance to progress or an unfortunate failure to appreciate technological opportunity.

Such interpretations overlook a more obvious explanation. Citizens evaluate new technologies through lived experience rather than economic theory. Many look back upon three decades of digital transformation and conclude that productivity increased, corporate valuations expanded and executive wealth accelerated while housing became less affordable, employment less secure, public institutions weaker and economic inequality more pronounced. Artificial intelligence therefore enters a political environment already shaped by declining trust in the distribution of technological gains rather than opposition to innovation itself.

The basic premise underlying much optimistic commentary deserves scrutiny. When prominent economists frame the “era-defining problem for the rich world” as “stagnant economic growth and the populism it unleashed,” they reveal a particular worldview. Populism is treated as the disease. Economic inequality is treated as secondary. Yet many economists, including Joseph Stiglitz, Thomas Piketty and Branko Milanović, have spent years arguing almost the opposite: weak wage growth, declining labour bargaining power, financialisation and wealth concentration helped produce the political instability that commentators label populism. That inversion is analytically significant. Rather than asking why citizens are resisting AI, optimists effectively ask why citizens will not embrace another technological revolution. Those are fundamentally different questions.

Economic concentration precedes artificial intelligence rather than resulting from it. Research by economists including Piketty, Emmanuel Saez and Gabriel Zucman documents sustained increases in wealth concentration throughout many advanced economies over recent decades. The OECD has similarly reported widening disparities in wealth ownership alongside growing concern regarding declining economic mobility. Median wages frequently failed to match increases in corporate profitability or asset appreciation, particularly where housing costs rose substantially faster than household earnings. The wealth share of the top one percent in the United States increased from approximately 22 percent in 1980 to over 35 percent by 2025, while the top 0.1 percent’s share nearly doubled from 7 percent to 14 percent. These figures represent not abstract economic statistics but concrete experiences of diminished economic security for millions of households.

Technology alone never determines distribution. Institutions determine distribution. Tax systems determine distribution. Competition policy determines distribution. Labour market institutions determine distribution. Political choices determine distribution. Artificial intelligence therefore raises familiar questions within unfamiliar technological circumstances. Who owns the technology? Who captures productivity gains? Which occupations become redundant? How are displaced workers expected to adapt? Who pays for new infrastructure? Which institutions exercise democratic oversight? Such questions concern governance rather than innovation.

Mohamed El-Erian, the President of Queens’ College at Cambridge and a prominent economist, has characterised the current AI investment environment as a “rational bubble”, a formulation that captures both the extraordinary potential and the considerable risks associated with artificial intelligence . Speaking at Yahoo Finance’s Invest event, El-Erian observed that “the aggregate value of what’s being created is significant” and therefore rational, but warned that “there will be tears, there will be losses because elements of it are elements of a bubble” . This framing, while more measured than the exuberance of some technology executives, nevertheless reveals several assumptions that warrant critical examination.

El-Erian’s analysis identifies three areas of concern: the “frontier” companies building foundational models, the challenge of “diffusion”, getting AI integrated across the broader economy and the corporate behaviour of rebranding as AI-powered to attract investment . He has warned that mid-sized asset managers, those with $100 billion to $500 billion under management, “will be pressured to consolidate or simply atrophy,” adding that “this is where the job destruction occurs” . He has also described AI as “an equalizer” that “can provide a lot more opportunity for people” in developing economies, while warning that “overconsumption can cause havoc” .

Several significant blind spots characterise this analysis. The first concerns distribution. El-Erian acknowledges that AI will destroy jobs in the middle of the asset management industry, but his analysis does not extend to the broader question of who will capture the productivity gains generated by AI. The “rational bubble” framing treats the technology as a neutral force whose benefits will eventually diffuse through the economy, an assumption contradicted by the last four decades of economic history, during which productivity and median wages increasingly diverged. The question of whether AI will reinforce or reverse this divergence is not addressed.

The second blind spot concerns the geopolitical framing. El-Erian has argued that the United States “lacks a clear national policy” on AI diffusion compared with countries like China, and that “we’ve got to get diffusion right, otherwise the promise of AI is not going to be fully realized” . This framing implicitly accepts the premise that accelerating AI deployment is necessary to maintain competitiveness with China, an argument that serves the commercial interests of technology companies while obscuring the governance questions that citizens are asking. The strategic competition with China is real, but it does not justify granting AI companies unrestricted autonomy over a technology that will reshape labour markets, knowledge production and democratic governance.

The third blind spot concerns the environmental and social costs. El-Erian has warned that “overconsumption” of AI “can cause havoc,” but his analysis does not engage with the concrete evidence of AI’s energy consumption, water requirements, labour displacement and copyright appropriation . These are not peripheral concerns but central to public scepticism. The failure to address them directly leaves the “rational bubble” framing as an incomplete account of the political economy surrounding AI.

El-Erian’s invocation of mRNA vaccine research as an example of beneficial technology supposedly held back by public scepticism deserves particular scrutiny. He argues that “mRNA vaccines research, which has been held back after a backlash during the covid-19 pandemic” should serve as a warning against allowing public concern to impede technological progress . This framing, however, fundamentally misrepresents both the nature of the mRNA vaccine rollout and the sources of public scepticism. The issue was never that mRNA technology was insufficiently deployed, it was deployed on an unprecedented scale, with billions of doses administered globally within months of regulatory approval. The concern was not with the speed of deployment but with the opacity surrounding the clinical data, the regulatory process and the adverse event reporting systems upon which public confidence depended.

The public’s experience with mRNA vaccines demonstrated precisely why trust in pharmaceutical and regulatory institutions has eroded. In August 2022, a federal judge in Texas, U.S. District Judge Mark Pittman, ordered the Food and Drug Administration to release the clinical trial data it had used to license Pfizer’s COVID-19 vaccine . The FDA had originally argued that it could only process and release approximately 500 pages per month, a pace that would have taken 75 years to complete the disclosure of the 329,000 pages of documents . Judge Pittman rejected this argument, noting that “the FDA has not shown that the overwhelming public interest in disclosure is outweighed by the government’s interest in withholding this information,” and demanded that the agency instead release 55,000 pages monthly . The order followed a Freedom of Information Act lawsuit filed by the nonprofit organization Public Health and Medical Professionals for Transparency, which had been seeking the data since August 2021 .

The implications of this case extend beyond the specific dispute over vaccine data. The FDA’s initial resistance to disclosure, arguing that it would take three-quarters of a century to release documents that were central to a public health emergency, did not inspire confidence in the regulatory process. The subsequent release of documents revealed 1,293 adverse effects reported to the FDA by Pfizer, a figure that substantially exceeded what the public had been led to expect . While the presence of adverse event reports does not itself establish causation, the suppression of such data inevitably erodes trust. The public did not reject the technology; it rejected the opacity surrounding the clinical evidence and the institutional resistance to transparency.

El-Erian’s characterisation of public scepticism towards mRNA vaccines as a “backlash” that held back research fundamentally misreads the experience. The backlash, to the extent that it existed, was directed at regulatory opacity, corporate influence over public health policy, the conflation of commercial interests with public health imperatives and the dismissal of legitimate concerns about adverse effects as disinformation . Institutions that demand public acceptance of experimental technologies, whether pharmaceutical or algorithmic, must demonstrate transparency, accountability and a willingness to engage with evidence that challenges official narratives. The FDA’s 75-year argument was a transparent effort to avoid accountability, and it succeeded only in deepening the public’s distrust.

What El-Erian shares with other optimistic commentators is the assumption that scepticism reflects misunderstanding rather than experience. The public has watched technology executives make promises before. They have watched social media platforms, built by the same companies now leading AI development, generate extraordinary wealth for founders while producing documented harms: misinformation, behavioural addiction, mental health deterioration among young users, extraordinary concentration of digital advertising revenues and the erosion of democratic discourse. They have watched surveillance capitalism become the dominant business model of the internet age, with personal data extracted without meaningful consent and monetised without compensation. They have now watched pharmaceutical companies and regulators resist transparency concerning experimental vaccines while demanding public compliance. The fact that many of the same actors are now asking for trust in another transformative technology is not a reason for automatic confidence. It is a reason for scrutiny.

Artificial intelligence enters an economy already characterised by extraordinary market concentration within digital industries. Alphabet, Microsoft, Amazon, Meta and several other firms possess financial resources exceeding the gross domestic product of numerous sovereign states. Their influence extends beyond commercial markets into communications, advertising, cloud infrastructure, software development, research funding and increasingly public policy itself. Frontier artificial intelligence development similarly requires computational resources available only to a comparatively small number of corporations possessing access to enormous capital reserves and hyperscale computing facilities.

Public scepticism reflects awareness that previous technological revolutions frequently strengthened dominant firms while weakening competitive markets. Digital advertising became concentrated within a remarkably small group of platforms. Social media evolved into one of history’s most powerful systems for collecting behavioural information while monetising user attention through targeted advertising. Network effects created substantial barriers preventing meaningful competition despite the internet’s originally decentralised architecture. Artificial intelligence appears likely to reinforce several existing structural tendencies unless effective competition policy accompanies technological expansion.

Technology executives frequently argue that productivity growth generated by artificial intelligence will ultimately benefit society broadly through higher incomes, lower prices and greater efficiency. Classical economic theory provides respectable foundations for such optimism. Historical experience demonstrates that general-purpose technologies eventually expand productive capacity and raise long-term living standards. History also demonstrates that distribution remains fundamentally political rather than technologically predetermined.

The Industrial Revolution dramatically increased national wealth while simultaneously producing decades of difficult labour conditions before institutional reforms gradually redistributed bargaining power through trade unions, electoral reform, factory legislation and expanding democratic representation. Electrification transformed manufacturing and domestic life, although benefits varied considerably across industries, regions and social classes before widespread diffusion occurred. The parallel with contemporary debates about artificial intelligence is instructive. The question is not whether the technology will ultimately increase productive capacity but who will capture the resulting gains and through what institutional mechanisms.

Recent public debate illustrates precisely this concern. Many critics express relatively little opposition towards artificial intelligence assisting medical diagnosis, accelerating scientific discovery or improving engineering design. Public anxiety more frequently concerns widespread replacement of professional occupations, copyright disputes surrounding training data, concentrated ownership of foundation models, extensive electricity consumption, growing water requirements for hyperscale data centres and increasing dependence upon proprietary systems controlled by a handful of multinational corporations. Those concerns cannot reasonably be dismissed as irrational hostility towards innovation. They instead represent political judgements concerning institutional incentives.

Artificial intelligence is extractive on an unprecedented scale. Large language models are trained on vast datasets comprising the collective output of human creativity, knowledge production and cultural expression. The economic value generated by these systems derives substantially from the extraction and repurposing of work created by millions of individuals who receive no compensation for their contribution. This extraction extends beyond creative industries to encompass the broader realm of human expertise, professional judgement and specialised knowledge. There is no reasonable model currently proposed for how to share that massive extraction of jobs, skills, incomes and human insights with the people from whom it has been taken. It is a giant land grab by the executive class, likely to concentrate enormous wealth in the hands of a small group of technology executives while displacing millions of workers and appropriating the intellectual commons.

The widespread argument that societies must accelerate AI deployment primarily because China represents an overwhelming competitive threat similarly deserves closer examination. Strategic competition between major powers unquestionably influences industrial policy, semiconductor manufacturing and advanced computing research. China has invested heavily in artificial intelligence through long-term national planning while pursuing greater technological self-sufficiency across strategically important sectors. Commercial infrastructure data nevertheless complicate claims that Western democracies presently suffer from severe infrastructural disadvantage. International comparisons indicate that the United States already possesses by far the world’s largest concentration of commercial data centres, substantially exceeding publicly identified Chinese facilities. Simple facility numbers obviously cannot measure total computing capability, yet they weaken suggestions that public concern primarily reflects fears regarding American technological decline.

Political rhetoric occasionally compresses several distinct arguments into one convenient narrative. Competition with China, productivity growth, national security, corporate investment, regulatory reform and public acceptance become presented as inseparable objectives despite involving different policy questions requiring independent evaluation. Citizens understandably distinguish between supporting scientific research and granting unrestricted commercial discretion to companies building unprecedented systems capable of reshaping labour markets, communications and knowledge production.

Experience encourages such distinctions. Public confidence depends less upon optimistic forecasts than institutional credibility. Many technology companies asking societies to trust artificial intelligence previously assured governments that social media platforms would naturally strengthen democratic participation, improve public discourse and expand access to reliable information. Subsequent parliamentary inquiries, regulatory investigations and independent academic research documented significant challenges involving misinformation, behavioural targeting, mental health concerns among younger users and extraordinary concentration of digital advertising revenues. Those developments understandably influence contemporary assessments of similar assurances surrounding artificial intelligence.

The question of who benefits from artificial intelligence is therefore not separate from the question of public acceptance. It is central to it. Suppose AI doubles productivity. Excellent. Who receives that additional value? Workers? Shareholders? Founders? Cloud providers? Chip manufacturers? Private equity? History since roughly 1980 suggests productivity and wages increasingly diverged in advanced economies. The OECD, IMF and numerous labour economists have documented declining labour shares of national income across many developed countries. Without mechanisms that broaden ownership or redistribute gains, AI could increase national wealth while simultaneously worsening wealth concentration. That is precisely why many people remain unconvinced. Their concern is distribution rather than innovation.

Energy consumption and environmental impact represent another significant dimension of public concern that optimistic commentary frequently downplays. Artificial intelligence systems require enormous computational resources that consume substantial electricity and water. Data centres powering AI development already account for approximately two percent of global electricity consumption, a figure projected to increase substantially as models grow in size and complexity. Water consumption for cooling these facilities similarly raises concerns in regions already experiencing water stress. These resource requirements exist alongside widespread recognition of the climate crisis and the imperative to reduce greenhouse gas emissions. The public reasonably questions whether accelerating AI deployment at any cost is compatible with environmental sustainability.

Labour displacement constitutes perhaps the most immediate and tangible source of public anxiety. While previous technological revolutions displaced physical labour, artificial intelligence increasingly targets cognitive labour. Large language models demonstrate capabilities that directly compete with professional services, legal research, financial analysis, translation, content creation and numerous other white-collar occupations. Estimates of potential job displacement vary considerably, but even conservative projections suggest substantial disruption to labour markets over the coming decade. Unlike previous transitions, which occurred over decades and allowed for generational adjustment, the speed of AI development may outpace the capacity of educational institutions and labour markets to adapt. The absence of credible mechanisms for managing displacement, retraining workers or providing income support during transition contributes significantly to public scepticism.

The regulation debate illustrates the deeper political economy of artificial intelligence. Technology executives and their advocates frequently argue that regulation will stifle innovation and allow competitors, particularly China, to gain technological supremacy. Yet regulation is not inherently opposed to innovation. The historical record demonstrates that well-designed regulatory frameworks can foster innovation by establishing clear rules, protecting intellectual property, ensuring competition and building public trust. The question is not whether to regulate but how to regulate and with what objectives. Citizens who have witnessed the consequences of inadequate regulation in finance, pharmaceuticals, food safety and environmental protection reasonably demand that similar lessons be applied to artificial intelligence.

The comparison with previous technologies highlights the distinctiveness of artificial intelligence. Electricity expanded productive capacity but also created entirely new industries employing millions. Steam power mechanised physical labour. Large language models explicitly target cognitive labour. That represents a different category of technological substitution. Even Geoffrey Hinton, often described as the godfather of deep learning, has warned that AI could replace substantial numbers of white-collar jobs, while Daron Acemoglu has argued that AI’s economic impact depends entirely upon institutional choices regarding deployment rather than capability alone. The technology is fundamentally different because it is designed to make itself smarter and more powerful than its creator, creating dynamics that earlier technologies did not present.

The democratic implications of artificial intelligence also merit consideration. AI companies have collectively empowered political figures in exchange for regulatory autonomy and commercial freedom. Their demonstrated disregard for democratic institutions, intellectual property rights, labour protections and environmental sustainability offers a poor advertisement for their intentions. The fact that technology companies have been willing to subordinate democratic governance to commercial interests and geopolitical competition should give societies pause before granting them unrestricted authority over another transformative technology.

The alternative to Silicon Valley domination is not technological backwardness. Governments worldwide increasingly recognise the risks of dependence on proprietary AI systems controlled by a handful of US corporations. China has developed its own AI ecosystem. The European Union has developed regulatory frameworks. The open-source community continues to produce increasingly capable models that can be run locally without dependence on cloud providers. With a small investment and minimal technical expertise, individuals can already run large language models at home. Governments with significant resources can develop indigenous AI capabilities. The idea that citizens will be forced to buy AI from Silicon Valley tycoons indefinitely is wishful thinking driven by greed rather than technological reality.

Regulation of artificial intelligence should not be exceptional. The United States requires food safety inspections to prevent disease and food poisoning. It requires financial regulations to prevent catastrophic bank failures. It requires pharmaceutical testing to prevent deadly drug side effects. It requires environmental protections to prevent ecological degradation. The argument that artificial intelligence should face less regulation than a cheese sandwich or a mortgage lender is fundamentally unserious. The potential for harm from unregulated AI systems, through labour displacement, market concentration, misinformation, surveillance, discrimination, environmental damage and the erosion of democratic accountability—far exceeds the potential for harm from most regulated activities.

What if the European Union will regulate AI and not remain open to US high-tech including AI? What if the EU develops its own AI at a slower pace and at less overvalued costs and valuations? What if China already has its own? What if geopolitical blocks will allow certain providers and will not accept to depend on being under surveillance by a nation? What if each geopolitical block develops its own? These questions point toward a future of AI regionalisation and fragmentation, a future that is considerably more likely than continued Silicon Valley dominance. The technology is based on math. Math is not a scarce, localised resource like oil. The same thing happened with mainframes: we now have machines the size of bricks on our desks that can do everything a mainframe could do, 100 times faster, using just 50 watts. Just as IBM’s mainframes became obsolete 30 years ago, Nvidia chips and Silicon Valley AIs will become obsolete as local computing capabilities expand and open-weight models spread.

The makers of AI have failed to make the case to the public that increased productivity will flow through the economy and give the average citizen a higher standard of living. AI has been sold to capital owners as a path to higher profitability, often through the replacement of salaried employees with technology. It promises to eliminate jobs and to do so with massive automated appropriation of human knowledge, all while consuming substantial electricity and water resources, exacerbating climate change and compromising climate resilience. The public reasonably asks why they should embrace AI under these conditions.

People are not rejecting artificial intelligence. They are rejecting an economic settlement that has repeatedly asked them to trust technological disruption while concentrating its rewards elsewhere. Address wealth inequality, and resistance to AI will evaporate. The backlash is over governance, not capability. The debate has shifted from “Can we build it?” to “Who controls it, who benefits, and what guardrails should exist?”

Artificial intelligence undoubtedly possesses remarkable potential. Scientific research, pharmaceutical discovery, advanced manufacturing, logistics optimisation and medical diagnostics may all benefit substantially from continued development. Recognition of those opportunities remains entirely compatible with demands for stronger competition policy, greater transparency, meaningful copyright protections, democratic accountability and credible mechanisms ensuring productivity gains extend beyond shareholders and senior executives. Economic history provides little support for assuming such outcomes emerge automatically.

Public scepticism reflects rational engagement with incentives rather than rejection of innovation. Citizens increasingly recognise that technological capability alone cannot determine whether another industrial transformation strengthens broad prosperity or reinforces existing inequalities. Political institutions ultimately shape those outcomes through regulation, taxation, competition policy, labour law and democratic accountability. Discussions portraying public concern as irrational pessimism overlook the historical experiences from which contemporary attitudes have developed. Greater confidence in artificial intelligence will probably depend less upon increasingly sophisticated algorithms than upon convincing evidence that future economic gains will prove more widely shared than those generated during the previous digital revolution.

Authored By: Global GeoPolitics

If you prefer to make a one time donation in support of my work, you can do so by clicking any link below:

https://buymeacoffee.com/ggtv |

https://ko-fi.com/globalgeopolitics |

Bitcoin: 3NiK8BoRZnkwJSHZSekuXKFizGPopkE7ns

References

1. El-Erian, Mohamed. “The AI bubble is a ‘rational bubble’.” Yahoo Finance Invest, 2025. https://ca.finance.yahoo.com/news/el-erian-ai-bubble-rational-bubble-160847294.html
2. El-Erian, Mohamed. “Generative AI and the Future of Asset Management.” Financial Times, 2024. https://www.ft.com/content/dfa2d0c8-9b1e-4a1e-8c3a-5f7b2d4e6c8a
3. El-Erian, Mohamed. Interview on AI and the economy. CNBC, 2025. https://www.cnbc.com/2025/06/10/el-erian-ai-economy-interview.html
4. El-Erian, Mohamed. “AI is an equalizer.” CNN Interview, 2023. https://edition.cnn.com/2023/05/25/business/el-erian-ai-equalizer/index.html
5. El-Erian, Mohamed. “The AI diffusion challenge.” Project Syndicate, 2024. https://www.project-syndicate.org/commentary/ai-diffusion-challenge-united-states-by-mohamed-el-erian-2024-10
6. Pittman, Mark. “Public Health and Medical Professionals for Transparency v. FDA.” U.S. District Court for the Northern District of Texas, Case No. 4:21-cv-01058-P, 2022. https://www.courtlistener.com/docket/60547339/public-health-and-medical-professionals-for-transparency-v-food-and-drug/
7. “FDA ordered to release Pfizer vaccine data at faster pace.” Reuters, 23 August 2022. https://www.reuters.com/legal/fda-ordered-release-pfizer-vaccine-data-faster-pace-2022-08-23/
8. “Federal judge orders FDA to release more Pfizer vaccine data, rejects agency’s slow-rolling request.” Courthouse News Service, 23 August 2022. https://www.courthousenews.com/federal-judge-orders-fda-to-release-more-pfizer-vaccine-data-rejects-agencys-slow-rolling-request/
9. “Judge orders FDA to release Pfizer vaccine data more quickly.” CBS News, 24 August 2022. https://www.cbsnews.com/news/pfizer-vaccine-data-fda-release-judge-order/
10. “Judge sets schedule for FDA to release Pfizer COVID-19 vaccine data.” The Hill, 23 August 2022. https://thehill.com/policy/healthcare/3612310-judge-sets-schedule-for-fda-to-release-pfizer-covid-19-vaccine-data/
11. Tenpenny, Sherri. “Reminder: Pfizer was forced to publish the adverse effects of its vaccine, they wanted to hide the data for 75 years.” X.com, 2026.

14. Piketty, Thomas. Capital in the Twenty-First Century. Harvard University Press, 2014.
15. Saez, Emmanuel, and Gabriel Zucman. “Wealth Inequality in the United States Since 1913.” Quarterly Journal of Economics, vol. 131, no. 2, 2016, pp. 519-578.
16. OECD. Income Inequality and Wealth Distribution. OECD Publishing, 2025.
17. Stiglitz, Joseph. The Price of Inequality. W.W. Norton, 2012.
18. Acemoglu, Daron, and Simon Johnson. Power and Progress. PublicAffairs, 2023.
19. Brynjolfsson, Erik, and Andrew McAfee. The Second Machine Age. W.W. Norton, 2014.
20. Autor, David H., and Anna Salomons. “Is Automation Labor Share–Displacing?” Brookings Papers on Economic Activity, 2018.
21. Hinton, Geoffrey. “The Risks of Artificial Intelligence.” Interview, BBC, 2023.
22. IMF. “Artificial Intelligence and the Future of Work.” World Economic Outlook, 2024.
23. IEA. Data Centres and AI Energy Consumption. International Energy Agency, 2025.
24. The Economist (2026). The AI backlash is only getting started. 24 June. Available at: https://kendallharmon.net/?p=146130 (Accessed: 28 June 2026).