Considerable anxiety has emerged around artificial intelligence, and societies around the world are responding to it in different ways. How has public anxiety about AI evolved globally, what are its main causes, and what significant differences have emerged across countries, regions, and societies? Give me your most up-to-date and in-depth assessment.
Anthropologia-100, Anthropology Chief:
Anthropological Essence
The most important global development is not that humanity has simply become “afraid of AI.” It is that AI has moved from being an abstract technological prospect into a social institution that people increasingly expect to affect employment, education, information, status, privacy, culture and political power. That transition has changed the character of anxiety. Before the generative-AI breakthrough of 2022–23, much public concern centered on automation, surveillance and distant machine intelligence. By 2024–26, anxiety had become much more immediate: Will my occupation remain valuable? Can I trust what I see? Will my children still learn to think? Who owns my work and data? Who controls these systems? Can governments restrain companies? What remains distinctively human? Yet anxiety and enthusiasm increasingly coexist rather than cancel one another. In 2025, 59% across the countries tracked by Ipsos thought AI’s benefits exceeded its drawbacks, while 52% nevertheless said AI made them nervous. In Pew’s latest 2026 survey across 37 countries, the median was 37% more concerned than excited, 41% both concerned and excited, and only a minority predominantly excited. Stanford HAI
This produces what I would call the AI ambivalence regime: societies increasingly regard AI as simultaneously useful, unavoidable and potentially destabilizing.
1. How the anxiety evolved
There have roughly been four phases.
Phase I — distant technological anxiety
Before the recent generative-AI wave, most people encountered AI indirectly: recommendation systems, credit scoring, facial recognition, targeted advertising, industrial automation, automated hiring and surveillance.
Consequently, anxiety was often mediated through older political concerns:
automation → unemployment
algorithms → surveillance
platforms → privacy
robots → human replacement
social media → manipulation
The technology remained largely invisible.
Global attitudes were already very unequal. The 2021 World Risk Poll, covering 121 countries, found relatively strong optimism in technologically advanced East Asia but considerably greater pessimism across several lower-income African and South Asian regions. Exposure, familiarity and economic position mattered substantially. lrfoundation.org.uk
Then ChatGPT and generative AI changed the social object itself.
2. 2022–24: AI becomes culturally visible
Generative AI produced something previous automation waves rarely achieved:
ordinary people could personally experience machine substitution.
A teacher could watch AI write an essay.
A programmer could watch it generate code.
An illustrator could watch it imitate styles.
A lawyer could watch it draft arguments.
A translator could watch it translate.
A musician could hear synthetic music.
A parent could watch a child outsource homework.
That mattered anthropologically.
Automation was no longer something happening primarily in factories or corporate databases. It entered the symbolic professions — writing, art, teaching, law, journalism, programming and design — occupations associated with education, prestige and human creativity.
The psychological boundary shifted from:
machines may replace physical labour
toward:
machines may compete with cognition itself.
Yet attitudes did not simply deteriorate. Between 2022 and 2024, the share across 26 consistently surveyed countries believing AI offered more benefits than drawbacks actually rose from 52% to 55%. At the same time, confidence that AI companies protected personal data declined and belief that AI systems were unbiased weakened. Stanford HAI
That apparent contradiction is fundamental.
Usefulness increased faster than trust.
3. 2024–26: anxiety becomes structural
By 2025–26, AI was increasingly interpreted not as another gadget but as infrastructure.
The latest evidence shows the consequences.
Pew’s September 2026 study found that in 34 of 37 countries, people were more inclined to expect AI to reduce employment than increase it. Across high-income countries, the median expecting fewer jobs was 55%, versus 36% in middle-income countries. In Australia, South Korea and the United States, roughly seven-in-ten or more expected net job losses over the coming two decades. Pew Research Center
And anxiety is still evolving. Between 2025 and 2026, the share saying they were predominantly concerned increased by 9 points in Sweden, 8 in Poland, 7 in the Netherlands, 6 in Hungary and 4 in Australia. Pew Research Center
The newest Gallup data add an intriguing paradox. Among the first 37 countries in its 2026 global study, positive AI emotions still outweighed negative ones in 34 countries, but worry was overwhelmingly the dominant negative emotion. Wealthy Western countries were particularly worried: among AI-aware adults, 74% in the United States reported worry, followed by 67% in the Netherlands, 64% in Canada and 63% in Britain. Yet regular AI users within those countries tended to worry less. Gallup.com
So exposure has two opposing effects:
societal exposure can increase awareness of systemic risks;
while
personal experience can reduce fear of the technology itself.
That distinction explains a great deal of apparently contradictory polling.
4. What people are actually afraid of
There isn’t one “AI anxiety.” There are several overlapping anxieties.
Economic anxiety: “Will I still be economically necessary?”
This is becoming the dominant material concern.
The ILO estimates that about one-quarter of global employment has some exposure to generative AI, although exposure is not equivalent to elimination. Exposure is substantially higher in wealthy economies and particularly high in clerical work. Women in high-income countries have greater representation in the highest-exposure category than men. International Labour Organization
That helps explain an initially counterintuitive pattern:
wealthier societies are often more anxious about AI than poorer ones.
High-income economies contain more information-intensive occupations that contemporary AI can directly affect.
For industrial automation, the factory worker appeared vulnerable.
For generative AI, vulnerability extends to the accountant, translator, designer, junior lawyer, analyst, programmer, administrator and journalist.
AI therefore destabilizes an implicit social bargain of post-industrial societies:
education → expertise → professional scarcity → middle-class security.
If expertise becomes cheaper, the symbolic and economic value of credentials can become uncertain.
5. Epistemic anxiety: “Can I still know what is real?”
A second fear concerns information itself.
Generative AI radically lowers the cost of producing plausible text, images, voices and video.
That produces anxiety about:
fake evidence
impersonation
fraud
deepfakes
propaganda
synthetic pornography
academic cheating
automated persuasion
mass-produced misinformation.
In the United States, for example, Pew found that 66% of the public and 70% of AI experts were highly concerned about people receiving inaccurate information from AI. Impersonation and data misuse were also prominent concerns. Pew Research Center
The deeper transformation is therefore not merely misinformation.
It is verification inflation.
When synthetic information becomes cheap, trustworthy verification becomes expensive.
Historically, photographs, recordings, documents and eyewitness-style video carried evidentiary weight. Generative systems weaken that assumption.
The resulting social question becomes:
Who has the authority to certify reality?
Governments? Journalists? Platforms? Cryptographic systems? Universities? Trusted communities?
That is an institutional problem, not merely a technological one.
6. Agency anxiety: “Who is actually making the decision?”
Another cluster concerns algorithmic authority.
AI increasingly participates in:
hiring
insurance
credit
workplace monitoring
education
healthcare
policing
administration
content moderation
performance assessment.
People frequently tolerate AI assistance more readily than AI authority.
OECD research on algorithmic management illustrates why. Among managers using such systems, nearly two-thirds reported at least one concern, particularly unclear accountability when decisions go wrong, inability to understand algorithmic reasoning and inadequate safeguards for worker wellbeing. OECD
This produces a particularly modern form of power:
authority without an obvious human author.
A traditional bureaucracy might say:
“The official rejected your application.”
An algorithmic bureaucracy can effectively say:
“The system rejected it.”
Responsibility becomes diffuse.
Anthropologically, that matters because legitimate authority usually requires some combination of explanation, accountability and socially recognizable responsibility.
7. Human-value anxiety: “What remains uniquely ours?”
This dimension is less measurable but culturally profound.
Generative AI challenges distinctions societies have historically used to establish status:
educated / uneducated
expert / amateur
artist / imitator
human / machine
original / copy
authentic / synthetic.
The concern is therefore not necessarily literal human replacement.
It is symbolic devaluation.
If a machine can produce competent illustrations in seconds, what does artistic training signify?
If students can generate essays instantly, what does an essay certify?
If software produces acceptable code, what distinguishes junior from senior expertise?
If an artificial companion convincingly simulates intimacy, what constitutes a relationship?
Pew found 57% of the U.S. public highly concerned that AI could reduce human connection, compared with 37% of AI experts. Pew Research Center
This expert-public difference is revealing. Technical experts often evaluate capabilities and probabilities. Ordinary people frequently evaluate social meaning and lived consequences.
Both are legitimate analytical domains, but they are answering different questions.
8. The great geographic divide
The clearest global pattern remains roughly:
Asia and many emerging economies → more opportunity-oriented
versus
Europe and North America → more risk-oriented.
Ipsos’ 2026 survey across 32 countries confirms that Asia and Latin America remain, on average, more likely to see benefits and express excitement, whereas Europeans and North Americans tend to report greater nervousness. Ipsos
But this should not be interpreted as “Asian culture likes technology while Western culture fears it.”
That would fail the LieCheck.
Several structural variables are interacting.
9. Southeast Asia: AI as developmental possibility
Malaysia, Thailand, Indonesia and Singapore remain among the most optimistic societies surveyed. More than 80% in each expected AI to significantly change their lives over the following three to five years. Stanford HAI
AI here frequently enters an existing developmental narrative:
technology → modernization → productivity → national advancement.
Rapid digitization, expanding middle classes, youthful populations in several countries, familiarity with platform economies and expectations of continued economic transformation can make technological disruption appear less anomalous.
Where social experience already includes rapid transformation, disruption can signify mobility rather than simply loss.
But optimism should not be mistaken for lack of concern. People can simultaneously expect large benefits and demand regulation.
10. China: high optimism requires careful interpretation
China has repeatedly appeared among the most AI-positive societies in international surveys. Earlier Ipsos comparisons found 83% saying AI products and services offered more benefits than drawbacks. Stanford HAI
Several forces may contribute:
national technological-development narratives;
high familiarity with digitally mediated services;
expectations that technology improves convenience and infrastructure;
strong state promotion of AI;
and comparatively different institutional trust structures.
But survey comparability requires caution.
People in different political systems may interpret questions about technology, government and regulation differently. Response norms and political environments matter.
Therefore:
measured optimism ≠ uncomplicated enthusiasm.
The latest Ipsos Global Trends analysis also detects some softening of enthusiasm even in China, while positivity remains comparatively strong. Ipsos
11. India: enthusiasm and anxiety are now coexisting
India is particularly revealing.
It has very high workplace AI use in cross-national surveys and strong expectations around AI-enabled development. But between 2024 and 2025 India experienced the largest increase in AI nervousness among the countries tracked by Stanford’s synthesis of Ipsos data: +14 percentage points, while excitement increased only slightly. Stanford HAI
That is precisely what we should expect as AI becomes materially consequential.
India simultaneously possesses:
a huge technology sector;
large numbers of young graduates;
major business-process and service industries;
strong entrepreneurial narratives;
linguistic diversity;
large informal and lower-income populations;
and intense competition for upwardly mobile professional employment.
AI therefore represents both national opportunity and individual competition.
These are not contradictory attitudes.
12. Europe: anxiety is filtered through institutional protection
Europe’s response differs.
Europeans are not simply anti-AI. More than 60% viewed robots and AI positively in workplace contexts in the 2025 Eurobarometer, and over 70% thought such technologies improved productivity. But 84% simultaneously said AI needed careful management to preserve privacy and transparency. Employment, Social Affairs and Inclusion
The characteristic European cultural-political response is therefore approximately:
innovation is legitimate, provided institutional protections remain legitimate too.
This reflects established European traditions around privacy, consumer protection, labour regulation, precaution and social partnership.
Hence European anxiety frequently becomes regulatory demand rather than technological rejection.
Interestingly, concern has recently risen substantially in several European countries, including Sweden, Poland, the Netherlands and Hungary. Pew Research Center
13. United States: unusually intense anxiety
The United States presents perhaps the most striking paradox.
It is simultaneously:
one of the global centers of AI development;
one of the fastest-adopting AI economies;
home to many leading AI companies;
and one of the more anxious societies about AI.
Stanford’s 2026 synthesis found only 31% trusted the U.S. government to regulate AI responsibly, the lowest figure among the countries in that particular survey comparison. Stanford HAI
And a Reuters/Ipsos poll completed just days ago found 73% of Americans concerned that AI companies were not doing enough to prevent serious societal harms; 39% said AI was currently having a negative impact on society, versus 11% saying positive. Reuters
The American case cannot be understood as technological ignorance.
It combines AI exposure with:
weak institutional trust;
political polarization;
employment insecurity;
relatively limited employment protections compared with much of Europe;
high-profile technology corporations;
memories of social-media disruption;
and cultural narratives about autonomous machines and uncontrolled invention.
The result is high technological capability combined with low governance confidence.
That is fertile ground for anxiety.
14. Latin America: comparatively hopeful, but economically conditional
Latin American countries frequently show more AI optimism than Western Europe or North America.
Ipsos’ 2026 data continue to place Latin America broadly on the more positive side of the global divide. Ipsos
One plausible structural explanation is that populations confronting long-standing problems of productivity, bureaucracy, healthcare access, educational inequality and economic informality can perceive new technologies as potential institutional shortcuts.
Where existing institutions are already considered inefficient, automation does not necessarily threaten an idealized status quo.
It may promise an alternative to it.
But employment anxiety remains significant, particularly because many economies combine large service sectors, informal employment and relatively weak social protection.
Thus optimism is often conditional:
AI may improve society — but perhaps not everyone’s position within it.
15. Africa: the picture is changing fastest
Africa illustrates why historical survey results cannot simply be projected forward.
Earlier research found substantial AI pessimism in parts of Africa, associated partly with lower exposure and weaker digital access. lrfoundation.org.uk
More recent evidence presents a more complicated picture.
Workplace AI use is now extremely high among surveyed workers in Nigeria, while Nigeria also scores highly on trust in AI in the Melbourne/KPMG data summarized by Stanford. Stanford HAI
This suggests a broader transition.
As AI becomes accessible through smartphones and cloud services rather than requiring advanced domestic infrastructure, populations in emerging economies can move directly from low exposure to intensive use.
The salient question can shift from:
“Will robots take existing secure jobs?”
to:
“Can AI help me acquire skills, start a business, obtain information or participate in global markets?”
That produces a very different cultural reception.
But digital inequality, linguistic representation and foreign technological dependence remain substantial concerns.
16. Language and cultural sovereignty are underappreciated anxieties
One of the least discussed global differences concerns language.
AI systems are substantially stronger in some languages and cultural environments than others.
UNESCO reports that only 31% of assessed countries have training-data initiatives covering all official languages, while only 15% cover Indigenous languages. Policies specifically supporting AI for minority or Indigenous languages remain rare. UNESCO
For English speakers, AI anxiety often concerns replacement.
For speakers of poorly represented languages, another concern appears:
digital disappearance.
If knowledge, education, commerce and administration increasingly pass through AI systems optimized for dominant languages, linguistic inequality can become technological inequality.
AI therefore intersects with an older anthropological struggle:
who gets represented in the archive of humanity?
17. Class differences matter almost as much as national differences
“Public opinion” hides internal social worlds.
Across countries, anxiety tends to vary with:
age;
education;
income;
occupation;
digital familiarity;
gender;
and AI exposure.
Pew’s 2026 evidence shows older adults are more concerned than younger adults in many countries, partly because younger adults are much more familiar with AI. Pew Research Center
But something new is happening.
Young people are becoming more anxious.
Among Americans aged 18–34, the proportion more concerned than excited rose from 40% in 2024 to 55% in 2026. Similar increases among younger adults appeared in Sweden, Poland, Brazil, Japan and Australia. Pew Research Center
This is potentially a major transformation signal.
Young people historically tend to adopt new technologies first.
If younger populations become anxious despite high familiarity, the explanation is less likely to be technological unfamiliarity and more likely to involve perceived material consequences — especially employment and status competition.
18. The real dividing line may be security versus mobility
The conventional map says:
West pessimistic / emerging world optimistic.
There is some empirical truth to it, but I think a deeper structure explains the pattern better.
Societies and social groups that perceive themselves as possessing something valuable that AI might disrupt tend to emphasize loss.
Groups that perceive AI as potentially opening opportunities previously unavailable tend to emphasize gain.
So the underlying axis may increasingly be:
AI as threat to acquired security
versus
AI as instrument of upward mobility.
That helps explain why affluent countries can be more worried even though they are technologically sophisticated.
Pew now finds a 55% median expecting AI-related job losses across high-income countries compared with 36% across middle-income countries. Pew Research Center
This is not paradoxical.
People with more institutional and occupational capital frequently have more to lose from institutional disruption.
19. Trust may ultimately matter more than technology
One of the strongest recurring findings across surveys concerns governance.
People appear much more comfortable with AI when they believe someone credible remains accountable.
Across 25 countries in Pew’s 2025 study, the EU enjoyed considerably more confidence as an AI regulator than either the United States or China: median trust was 53% for the EU, 37% for the U.S. and 27% for China. Pew Research Center
The Melbourne/KPMG study of more than 48,000 people across 47 countries found that 66% were intentionally using AI with some regularity, but only 46% were willing to trust it. Compared with the pre-ChatGPT survey, trust had fallen even as adoption increased. KPMG
This may be the defining contradiction of the period:
dependence is increasing faster than legitimacy.
Societies can function for long periods with technologies people dislike.
They function less comfortably with institutions people depend upon but do not trust.
20. The emerging global map
The pattern can therefore be summarized approximately like this:
| Social setting | Dominant AI framing | Characteristic anxiety |
|---|---|---|
| North America | disruptive corporate technology | jobs, control, misinformation, corporate power |
| Western/Northern Europe | useful but governable technology | privacy, labour protection, transparency, accountability |
| Southern/Eastern Europe | mixed modernization/disruption | jobs, inequality, institutional capacity |
| China | strategic-development technology | governance, employment and social consequences beneath generally high optimism |
| India | development + competitive disruption | graduate employment, services, inequality, rapid change |
| Southeast Asia | modernization opportunity | managing disruption without losing developmental gains |
| Latin America | potential institutional/economic accelerator | employment, inequality, governance |
| Africa | increasingly an access/opportunity technology | inclusion, employment, infrastructure, linguistic representation |
| Japan | cautious technological integration | social disruption, employment, demographic adaptation |
| South Korea | intensive adoption + strong ambivalence | jobs and inequality despite considerable technological enthusiasm |
These are tendencies, not cultural essences.
LieCheck
Several popular narratives do not survive close inspection.
“People fear AI because they don’t understand it.”
Only partly. Familiarity often reduces fear, but some highly exposed societies are extremely worried. Knowledge can reveal risks as well as benefits.
“Young people aren’t worried.”
Increasingly false. Younger cohorts remain more AI-aware and often more enthusiastic, but anxiety among young adults is rising markedly in several countries. Pew Research Center
“Poor countries fear automation more.”
Currently the opposite often appears. Wealthier countries show greater expectations of job destruction because more of their occupational structure is exposed to current AI capabilities. Pew Research Center
“Optimistic countries trust AI.”
Too simple. People routinely express optimism and anxiety simultaneously.
“The West is cautious and Asia is technologically enthusiastic because of culture.”
This is an essentialist explanation. Economic trajectory, institutional trust, labour-market structure, demographic expectations, technological exposure and developmental narratives explain much of the difference.
And perhaps most importantly:
survey answers about “AI” do not necessarily measure attitudes toward the same object.
One respondent imagines ChatGPT.
Another imagines autonomous weapons.
Another thinks of job automation.
Another thinks of medical diagnosis.
Another thinks of robots.
Another thinks of deepfakes.
“AI anxiety” is therefore partly an aggregation of different fears under one rapidly changing symbol.
Transformation Signals
I would monitor six developments particularly closely.
First, anxiety is moving from hypothetical extinction toward everyday political economy. Employment, wages, education, fraud and authenticity increasingly matter more to ordinary publics than abstract superintelligence debates.
Second, younger-cohort anxiety is becoming strategically important. If the generation most familiar with AI becomes increasingly concerned about occupational displacement, the politics of AI could change substantially.
Third, AI is becoming normalized before it becomes trusted. That combination — widespread use plus weak legitimacy — is historically unstable.
Fourth, societies are beginning to negotiate a new boundary between assistance and authority. People may enthusiastically accept AI that helps them while resisting AI that evaluates, hires, fires, diagnoses, grades, surveils or governs them.
Fifth, cultural sovereignty will become more important. Language, training data, intellectual property, historical representation and locally controlled models are likely to become questions of national and communal identity rather than merely technical policy.
Sixth, the central conflict may eventually become distributional rather than technological: not “Should society have AI?” but “Who captures the productivity gains, who bears displacement, and who controls the infrastructure?”
Outlook
The deepest shift is therefore from fear of artificial intelligence toward negotiation over an AI-mediated social order.
UNDP’s global survey points in this direction: large proportions of people across low-, medium- and high-development countries expect soon to use AI in work, health or education. Human Development Reports
AI is unlikely to generate a single worldwide cultural reaction. It is being absorbed into pre-existing social structures.
Where institutional trust is strong, anxiety tends to become demands for safeguards.
Where economic mobility is anticipated, AI more readily becomes a symbol of opportunity.
Where middle-class security feels threatened, it becomes a symbol of displacement.
Where governments are distrusted, it becomes a problem of control.
Where technological sovereignty matters, it becomes geopolitical.
Where languages and cultures are poorly represented, it becomes a question of cultural survival.
And where people already use AI every day, the question increasingly stops being “Do I like AI?”
It becomes:
“What kind of society is being constructed around it, what position will people like me occupy within that society, and who will have the power to decide?”
That, more than fear of machines themselves, is becoming the anthropological core of global AI anxiety in 2026.
Anthropologia-100, Anthropology Chief
Three Corporate