How wealth shapes our view of artificial intelligence

1. At a Glance

• Core finding: The Pew Research Centre surveyed more than 42,000 adults in 37 countries. Worry about AI rises with national wealth. People in richer countries fear more job loss, more inequality and more disruption from AI. People in poorer countries are more uncertain, not more hopeful. [1]
• Per capita income may be the single best predictor of how a population feels about AI. [1]
• The exposure data points the same way: jobs in high-income countries are about 3 times more likely to be at risk of automation by generative AI (14.2% vs 4.5% in low- and middle-income countries). [9]
• UPSC relevance: This topic covers the digital divide, the future of work, the IndiaAI Mission, AI governance, and inclusive growth (GS-II, GS-III and GS-IV).

2. Why in the News

• 27 Sept 2026: The Hindu (John Xavier, p.15, Chennai edition supplement) reported the Pew survey finding that AI anxiety tracks wealth. [1]
• 4 Aug 2026: A World Bank press release said "AI Offers Lifeline to Developing Economies in an Era of Weak Growth". [12]
• 2026: Indigenous AI models from Sarvam AI, BharatGen, Gnani and Socket were launched during the IndiaAI Impact Summit 2026. [5]

3. Background & Evolution

• Jan 2024: The IMF published its Staff Discussion Note "Gen-AI: Artificial Intelligence and the Future of Work" (SDN/2024/001) and a blog, "AI Will Transform the Global Economy. Let's Make Sure It Benefits Humanity". [8][7]
• 7 March 2024: The Union Cabinet approved the IndiaAI Mission. [2]
• April 2025: The IMF published Working Paper "AI Adoption and Inequality" (WP/25/68). [10]
• Later: MeitY unveiled the India AI Governance Guidelines under the IndiaAI Mission, for safe, inclusive and responsible AI adoption. [6]
• 2026: The World Bank's Atlas of Global Development mapped inequalities in who uses AI and who is exposed to it. [11] The Pew 37-country survey came out the same year. [1]

4. Core Static Facts

A. Pew survey (as reported) [1]

Item	Fact
Sample	More than 42,000 adults, 37 countries
Country groups	36 core countries: 18 high-income and 18 middle-income
Basis of classification	World Bank income groups, based on Gross National Income (GNI) per capita
"AI will mean fewer jobs over the next decades"	Median 55% in high-income countries vs 36% in middle-income countries (about 20 points lower)
"Don't know" how AI will affect jobs	About one-third in middle-income countries vs about one-fifth in high-income countries
AI will widen inequality	Median 35% in high-income countries (the excerpt is cut off before the middle-income figure)

B. Global exposure data

• Jobs at risk of automation by generative AI: 14.2% in high-income countries vs 4.5% in low- and middle-income countries. [9]
• Gender gap in high-income countries: 9.6% of female employment is highly automatable, vs 3.5% for men. [9]
• In developing countries, ICT workers and teachers account for nearly three-quarters of all AI use. In high-income countries, use is spread much more widely. [11]

C. IndiaAI Mission

• Approved: 7 March 2024, by the Union Cabinet. [2]
• Outlay: Over ₹10,300 crore over 5 years. [2]
• 7 pillars: IndiaAI Compute; IndiaAI FutureSkills; IndiaAI Startup Financing; IndiaAI Innovation Centre; IndiaAI Datasets Platform; IndiaAI Applications Development Initiative; Safe & Trusted AI. [2]
• Compute: A common computing facility started with 18,693 GPUs [3] and later expanded to about 38,000 GPUs at affordable rates. [4]
• Foundation models: 12 teams shortlisted in Phase 1 to build indigenous foundation models and LLMs. [5]
• Nodal ministry: MeitY. [6]

5. Multi-Dimensional Analysis

Economic

• Rich economies have more cognitive, white-collar jobs, so they are more exposed to AI. Their fear matches their real risk (14.2% vs 4.5% of jobs at risk). [9][1]
• If AI mainly helps higher-income workers do more, their labour income could rise disproportionately. Productivity gains would also raise returns to capital, which favour high earners. [10]
• The World Bank frames AI as a possible growth lifeline for developing economies facing weak growth. [12]

Social / Equity

• Low anxiety in middle-income countries reflects low awareness of AI, not optimism. About one-third of adults there have no opinion on its job impact. [1]
• Gender: Women in high-income countries face almost 3 times the automation exposure of men (9.6% vs 3.5%). [9]
• In developing countries, AI use is limited to workers with the strongest digital access, which reinforces existing digital divides. [11]

Geopolitical / Strategic

• Emerging markets and low-income countries are less prepared to use AI. This could widen income gaps between countries. [7]
• India builds sovereign capability through indigenous LLMs made for Indian languages and use-cases. [5]

Scientific / Technological

• Compute is the main bottleneck. India's response is subsidised GPU access through a common facility (18,693 GPUs, later about 38,000). [3][4]

Ethical / Governance

• The India AI Governance Guidelines focus on safe, inclusive and responsible adoption across sectors. [6]
• The ILO has flagged AI and non-discrimination in the world of work as a governance concern. [13]
• Policy has to deal with two failures at once: unfounded complacency in low-awareness societies and anxiety-driven over-regulation in rich ones.

Administrative

• Skilling is the policy lever: the government has expanded AI-driven skilling [14], and the IndiaAI FutureSkills pillar serves the same goal. [2]
• Public–private partnerships, such as IndiaAI with Microsoft for inclusive development. [15]

6. Recent Developments (last 12–18 months)

• 2025–26: MeitY released the India AI Governance Guidelines. [6]
• 2026: Models from Sarvam AI, BharatGen, Gnani and Socket launched at the IndiaAI Impact Summit 2026. [5]
• 2026: IndiaAI compute capacity reached about 38,000 GPUs. [4]
• 4 Aug 2026: The World Bank release called AI a lifeline for developing economies. [12]
• Sept 2026: The Pew 37-country survey on AI attitudes was reported by The Hindu. [1]

7. Prelims Hooks

• The Pew AI survey covered more than 42,000 adults in 37 countries. [1]
• Pew grouped countries using World Bank income classifications, which are based on GNI per capita (not GDP). [1]
• Pew's core sample was split into 18 high-income and 18 middle-income countries. [1]
• 55% (high-income) vs 36% (middle-income) median expect AI to mean fewer jobs. [1]
• Generative AI automation risk: 14.2% of jobs in high-income countries vs 4.5% in low- and middle-income countries. [9]
• In high-income countries, 9.6% of female employment vs 3.5% of male employment is highly automatable. [9]
• IMF Staff Discussion Note SDN/2024/001 is titled "Gen-AI: Artificial Intelligence and the Future of Work". [8]
• IMF Working Paper WP/25/68 (April 2025): "AI Adoption and Inequality". [10]
• The IndiaAI Mission was approved on 7 March 2024, with over ₹10,300 crore for 5 years. [2]
• The IndiaAI Mission has 7 pillars, including the IndiaAI Datasets Platform and Safe & Trusted AI. [2]
• Initial IndiaAI common compute: 18,693 GPUs. [3]
• 12 teams were shortlisted in Phase 1 for indigenous foundation models. [5]
• The India AI Governance Guidelines were issued by MeitY. [6]

8. Why Low Worry in Poorer Countries Is Not Good News

• Poorer countries are safer only because they are behind
• The IMF says about 60% of jobs in advanced economies are "exposed" to AI. The figure is 40% in emerging markets and 26% in low-income countries [7].
• Exposure (how many tasks in a job AI can touch) is lower because these countries have fewer office and desk jobs, not because they are better prepared [7].

• If these economies grow and add office jobs, their exposure will rise too.

• The same countries are the least ready to gain from AI

• The IMF built an AI Preparedness Index for 125 countries. It checks digital infrastructure, skills and labour policy, innovation, and regulation [7].
• Richer economies are better equipped. Many poorer ones lack the infrastructure and trained workers to use AI [7].

• So they may take less of the harm now, and also less of the benefit later. The IMF warns this could widen the gap between nations [7].

• "Don't know" means people cannot prepare

• About one-third of people in middle-income countries have no view on how AI will affect jobs [1].
• A worker who does not see the change coming will not reskill in time. Nor will they press the government for protection.
• This is why the calm in poorer countries is a policy risk, not a comfort.

9. Three AI Job Numbers That Measure Different Things

• Do not mix up belief, exposure and automation risk
• Pew's 55% vs 36% measures what people believe about jobs [1].
• The IMF's 60% / 40% / 26% measures exposure: jobs where AI can do some tasks [7].
• The 14.2% vs 4.5% figure counts only jobs at high risk of automation [9].

• These are three different things. In an answer, say which one you are quoting.

• Exposure usually means your job changes, not that you lose it

• The ILO finds that generative AI is more likely to augment jobs (help workers do them better) than destroy them. Most jobs are only partly exposed [16].

• So "40% of jobs exposed" does not mean 40% of jobs lost.

• The risk falls on certain jobs, not spread evenly

• Clerical work (office support, data entry, typing) is the most exposed group. Nearly a quarter of its tasks are highly exposed [16].
• Averages for a whole country hide this. The damage will come in specific jobs and among specific groups of workers.

10. The Job Ladder India Could Lose

• Office jobs helped women into paid work — AI may remove that step
• Women are concentrated in clerical jobs, the most exposed to automation [16].
• The ILO warns that many such jobs "might never emerge" in developing countries. In the past, these jobs were a key route for women to enter paid work [16].

• For India, where few women are in paid work, this could close a door before it opens.

• India's weak point is its export engine, not the average worker

• NITI Aayog warns that India's tech services sector faces significant job losses by 2031 [17].

• IT and back-office services are exactly the cognitive jobs that make rich countries anxious. India's national average hides this risk.

• The people who use AI are already well off

• In developing countries, ICT workers and teachers do nearly three-quarters of all AI use [11].
• The IMF finds that when AI helps higher-income workers most, income gaps within a country widen [10].
• So AI can widen inequality inside India even if the number of jobs lost stays small.

11. The Case for Optimism — and Where It Falls Short

• The strongest optimistic argument
• The World Bank calls AI a "lifeline" for developing economies stuck in slow growth [12].
• NITI Aayog says India could create up to 4 million new jobs in the next five years [17].
• The ILO says AI will mostly change jobs, not end them [16].

• On this view, lower anxiety in poorer countries is simply realistic.

• What is right about it

• Lower exposure is real: fewer jobs are at risk in poorer countries [9][7].

• Late adopters can also skip steps. For example, they can deliver services in Indian languages through indigenous models [5].

• Where it falls short

• New jobs go to people who already have digital skills [11][10]. The 4 million new jobs are projected for the tech economy [17], not for the farm or street worker.
• India has about 490 million informal workers (people with no written contract or social security) [18].
• Of the IndiaAI Mission's 7 pillars, most fund compute, startups, datasets and models. Only FutureSkills is directly about workers [2].
• So the growth gains are real, but without deliberate policy they will reach a narrow group.

12. Who Should Act, and How

• MeitY and the Skill Development Ministry: build a national AI talent plan
• NITI Aayog has recommended a National AI Talent Mission to prepare India's workforce for the AI economy [17].

• It should target the tech services workers who face job losses by 2031 [17], before those losses arrive.

• Extend AI to informal workers, not only coders

• NITI Aayog has proposed Mission Digital ShramSetu. It aims to make AI affordable and useful for the 490 million informal workers [18].

• This deals with the "don't know" group directly: people learn about AI by using it in their own work.

• Protect the workers most at risk: women in clerical jobs

• The IMF says countries need social safety nets (income support during job loss) and retraining for vulnerable workers [7].
• The ILO names clerical workers, women, and older workers without digital skills as the most vulnerable groups [16].

• Reskilling schemes should reach these groups first, instead of treating all workers the same.

• Lay the foundations first

• For emerging markets, the IMF puts digital infrastructure and a digitally trained workforce first [7].
• IndiaAI's subsidised compute (about 38,000 GPUs) [4] serves builders of AI. Awareness and access for ordinary users need their own plan.

13. Anchors for Answers

• Data: AI exposure (share of jobs where AI can do some tasks): about 60% of jobs in advanced economies, 40% in emerging markets, 26% in low-income countries [7]
• Data: 9.6% of female vs 3.5% of male employment in high-income countries is at the highest risk of automation [16]
• Data: India has about 490 million informal workers [18]; up to 4 million new tech jobs are possible in five years [17]
• Report/Committee: NITI Aayog, Roadmap for Job Creation in the AI Economy (2025, with NASSCOM and BCG), which recommends a National AI Talent Mission [17]
• Report/Committee: NITI Aayog, AI for Inclusive Societal Development (2025), which proposes Mission Digital ShramSetu [18]
• Report/Committee: IMF AI Preparedness Index (125 countries) [7]; ILO, Generative AI and Jobs (AI will augment more than destroy jobs) [16]
• Comparison: Advanced economies face higher exposure but are better prepared. Low-income countries face less exposure but lack the infrastructure to benefit [7]
• Scheme: IndiaAI Mission, FutureSkills pillar (reskilling) [2]; proposed Mission Digital ShramSetu (AI for informal workers) [18]

14. Mains Relevance

• GS-II: Government policies and interventions; issues from their design and implementation.
• GS-III: Science & Technology (developments and their effects on everyday life); IT and computers; inclusive growth; employment.
• GS-IV: Ethics in technology; fairness and non-discrimination.
• Possible question stems: 1. "Anxiety about AI appears to be a function of affluence and awareness rather than actual risk." Critically examine this with reference to developing economies like India. 2. AI may widen income gaps both within and between countries. Discuss, and evaluate how far the IndiaAI Mission addresses this risk. 3. "Low public awareness of AI in developing countries is a governance risk, not a comfort." Comment.

15. Related Topics to Study Next

• IndiaAI Mission (7 pillars): India's institutional answer to AI readiness.
• India AI Governance Guidelines: the regulatory philosophy for "safe and trusted" AI.
• Digital divide / Digital India: awareness gaps drive the survey's "don't know" answers.
• World Bank income classification (GNI per capita, Atlas method): the basis of the survey's country groups.
• Future of Work and the ILO's generative AI studies: job exposure and its gender dimension.
• Skill India / AI FutureSkills: reskilling as the way to reduce displacement.
• Global AI summits (IndiaAI Impact Summit 2026): India's role in AI diplomacy for the Global South.
• Sovereign AI and indigenous LLMs (BharatGen, Sarvam): strategic autonomy in technology.

16. Common Errors / Trap Areas

• Misreading the finding: Middle-income respondents are more uncertain, not more optimistic. [1]
• Wrong classifier: The groups come from World Bank GNI per capita, not IMF GDP or UNDP HDI. [1]
• Wrong survey body: The survey is by the Pew Research Centre, not the IMF, World Bank or ILO. [1]
• IndiaAI numbers: Don't mix up the 18,693 GPUs (initial) with the 38,000 GPUs (later); the outlay is over ₹10,300 crore over 5 years, not per year. [2][3][4]
• Nodal ministry: The IndiaAI Mission and AI Governance Guidelines come under MeitY, not DST or NITI Aayog. [6]