UN warns AI race could widen global inequality as countries struggle to keep pace

13 Min Read
UN Secretary-General Calls For Global Unity in AI Development Amid ...

Asia Pacific's AI boom leaves billions outside

Only about 14 percent of people in Asia Pacific actively use artificial intelligence tools, leaving roughly 3.7 billion people on the sidelines, even though the region accounts for more than half of global AI users. A United Nations Development Programme assessment warns that this gap between rapid technological progress and limited access could deepen inequality between countries.

Contents
  1. Asia Pacific's AI boom leaves billions outside
  2. What separates countries that can benefit?
  3. Large growth prospects meet fragile livelihoods
  4. Women and young workers face particular pressure
  5. Who gets represented in AI decisions?
  6. Public services show what useful AI can deliver
  7. Rules and security are struggling to catch up
  8. A small group holds much of the investment and influence
  9. Investment priorities depend on each country's starting point
  10. Key Points

The report, The Next Great Divergence, identifies a central divide: whether countries have the skills, computing power, infrastructure and institutions needed to benefit from AI. Adoption is happening in months, not decades. Many governments are still building the foundations needed to use the technology while protecting people whose jobs and livelihoods could change.

The warning comes after five decades in which many countries with lower incomes moved closer to wealthier economies through trade, technology and development. UNDP argues that AI could put those gains at risk if countries with greater resources pull ahead while others struggle to participate.

Asia Pacific sits at the center of this transition. It is home to more than half the world's population, yet one in four residents remains offline. In South Asia, women are up to 40 percent less likely than men to own a smartphone. Access to an AI service therefore begins with something much simpler: having a device and an internet connection.

The region also faces urgent needs beyond technology. About 1.6 billion people cannot afford a healthy diet, 27 million youths remain illiterate, and around 200 million people live in extreme poverty. These figures help explain why the question is bigger than who develops the most advanced AI. It is also about whether technological gains reach people already facing disadvantage.

Advertisement

What separates countries that can benefit?

UNDP describes capability as the main fault line in the AI era. In practical terms, capability means reliable and affordable electricity, resilient internet connections, useful data, access to cloud services and computing power, skilled workers, and institutions able to write and enforce rules.

Computing power is the capacity needed to run AI systems or adapt them for particular uses. Cloud services provide access to computing resources without requiring every organisation to build its own facilities. Countries need access to these resources alongside people who understand how to use them.

Some economies are advancing quickly across these areas. Others are still working on basic digital access and literacy. The gap affects how readily hospitals, schools, farms, financial institutions and public agencies can turn AI into useful services.

Philip Schellekens, Chief Economist for UNDP's Asia and the Pacific Regional Bureau, said the technology could reverse a long period of narrowing differences between countries.

We think that AI is heralding a new era of rising inequality between countries, following years of convergence in the last 50 years.

Schellekens also warned that rising inequality could have consequences beyond incomes and employment, including for security and migration.

If inequality continues to rise, the spillover effects of that in terms of the security agenda, in terms of undocumented forms of migration, will also become more daunting.

Kanni Wignaraja, UN Assistant Secretary General and UNDP's regional director for Asia and the Pacific, described how the speed of technological change compares with countries' readiness.

AI is racing ahead, and many countries are still at the starting line.

Large growth prospects meet fragile livelihoods

China accounts for nearly 70 percent of global AI patents, while more than 3,100 newly funded AI companies have emerged across six economies in Asia Pacific. These figures show the scale of the region's innovation alongside the large population still outside active AI use.

Advertisement

Modeling in the UNDP report suggests AI could lift annual GDP growth in Asia Pacific by around 2 percentage points. GDP measures the value of goods and services an economy produces. An increase in its growth rate means the economy expands faster each year.

The report also points to potential productivity gains of up to 5 percent in sectors such as health and finance. Higher productivity means getting more useful work done with the resources available. ASEAN economies alone could gain nearly $1 trillion in additional GDP over the next decade if adoption is broad and inclusive.

About 1.3 billion workers in the region, however, are in informal employment, with little protection if their jobs change or disappear. Nearly 770 million women are outside the labor force. The potential for faster growth therefore sits beside a large need for support, training and access to new opportunities.

Garment manufacturing hubs could face faster adoption of robotics, affecting industries that often employ women. Business process outsourcing, call centers and back office services in the Philippines, India and Bangladesh face competition from increasingly capable AI systems. These businesses carry out tasks such as handling customer inquiries and administrative work, many of which involve routine processes.

UNDP warns that disruption could outweigh gains in the near term for many workers unless governments support retraining, help people through job transitions and encourage new employment. One in four companies surveyed expects job losses linked to AI automation, indicating that some employers plan to substitute technology for tasks currently performed by people.

Women and young workers face particular pressure

Jobs held by women are nearly twice as exposed to automation as those held by men, according to the report's labor analysis. Women are overrepresented in roles involving routine, predictable tasks, which makes their work more susceptible to technological change.

Young workers face a different pressure. In several economies, employment is already declining in roles with high exposure to AI, especially among people aged 22 to 25. Losing these opportunities can weaken the early stages of a career, when people gain experience and build skills.

Advertisement

Basic digital skills remain a barrier to adapting. Only about one in four urban residents, and fewer than one in five rural residents, can complete simple spreadsheet tasks. Workers who lack these foundations have more to learn before they can use new tools confidently in their jobs.

Training therefore needs to cover both basic digital abilities and the practical use of AI in particular sectors. UNDP's priorities include learning programs focused on women and young people, with training connected to health, agriculture, education and small businesses.

Social protection is another part of the response. The proposed measures include job matching, wage insurance and incentives for firms to create new roles as automation spreads. Wage insurance is support intended to cushion an income loss during a job transition. For workers with little existing protection, these measures can help connect technological change with a viable next job.

Who gets represented in AI decisions?

Rural and indigenous communities, along with many women entrepreneurs, are often underrepresented in the datasets used to train machine learning models. Machine learning systems identify patterns in data and use those patterns to make predictions or guide decisions. Missing or unbalanced data can carry existing disadvantages into those decisions.

The report cites AI credit scoring models trained mainly on information about urban male borrowers. Those models can misclassify women and rural farmers as higher risk. A credit score helps a lender judge whether someone is likely to repay a loan, so an inaccurate assessment can restrict access to finance.

Women entrepreneurs and farmers may then find it harder to obtain money for their businesses or livelihoods. This connects a technical problem, the composition of training data, with an everyday economic consequence.

Many countries also depend on imported AI models that do not adequately reflect local languages, dialects or cultural settings. That can reduce accuracy and trust, particularly when the systems are used in health or finance, where errors can have serious consequences.

Local language support requires both data and computing resources, which are scarce in many countries with lower incomes. The report's proposed response includes representative data, support for local languages and dialects, open datasets, and open source projects designed around public needs.

Advertisement

Public services show what useful AI can deliver

Bangkok's Traffy Fondue platform has handled nearly 600,000 citizen reports, helping city teams respond to potholes, rubbish and other local problems more quickly. The example connects digital systems with visible improvements in the places people live.

Singapore's Moments of Life service reduced paperwork for new parents from roughly two hours to about fifteen minutes. Using those figures, the task takes about one eighth of the previous time, saving roughly one hour and forty five minutes.

Beijing is using digital twins to help planners simulate floods and traffic flows when making infrastructure decisions. A digital twin is a virtual representation that mirrors a physical system in real time. It lets planners examine how a system might respond to changing conditions.

Health services can use AI pattern recognition to speed the diagnosis of conditions such as tuberculosis from medical images. Education tools can adapt to students' needs and provide tutoring in remote schools. Agricultural systems can examine soil, weather and pest data to advise farmers on planting and fertilizer.

Food systems, disaster relief and social protection programs can also benefit from analysis of information as conditions change. That can help direct resources to the places and people that need them. The public service examples illustrate gains in responsiveness, reduced administrative work and practical value when technology is deployed with care.

Expanding the infrastructure behind these services brings another challenge. Rapid growth in data centers increases energy and water demand. Countries need to expand cloud and AI capacity while avoiding excessive pressure on power grids and water supplies, especially where climate goals are already under strain.

Rules and security are struggling to catch up

Only a limited number of countries have comprehensive AI regulations in place. The report calls for clearer accountability, testing standards, systems that can be audited, and privacy protections as AI becomes more widely used.

By 2027, more than 40 percent of global data breaches related to AI may stem from misuse of generative AI, according to the report. Generative AI produces material such as text, images and audio. Its ability to create convincing content can support useful work and also give criminals new tools.

Advertisement

A system that helps a teacher draft lesson plans can also be used to accelerate phishing attacks, create deepfake fraud or write malware. Phishing uses deceptive messages to obtain information or access. Deepfakes imitate a person's appearance or voice, while malware is software designed to cause harm or gain unauthorised access.

Governments need credible safeguards that people can understand and institutions can enforce. The priorities include transparency, safety evaluations and clear ways to seek a remedy when an AI system causes harm.

For uses with high risks in public and private organisations, the proposed measures include privacy laws, assessments of possible effects and independent audits. Audits allow a system's operation and decisions to be checked. Lenders should also validate AI credit scoring systems against discrimination, connecting governance directly with the problems facing women and rural borrowers.

A small group holds much of the investment and influence

A UN trade and development analysis finds that just 100 companies, mainly in the United States and China, account for about 40 percent of private investment in AI research and development. At the same time, 118 countries, mostly in the Global South, are absent from international discussions on AI governance.

The global AI market is projected to reach $4.8 trillion by 2033. With investment concentrated and many countries missing from discussions about rules, a small group of firms and governments could shape the infrastructure and standards on which others depend.

UNDP's report sets out the relationship between national preparedness and the distribution of benefits in these words:

The central fault line in the AI era is capability. Countries that invest in skills, computing power and sound governance systems will benefit, others risk being left far behind.

UN trade and development experts propose a shared global facility to provide more equitable access to computing power and AI tools. Pooled infrastructure could help researchers, regulators and small enterprises in countries with lower incomes gain access to resources they otherwise lack.

Public disclosure standards for AI systems could also improve transparency and accountability across borders. The proposal draws on the idea of environmental and governance reporting, giving a more structured way to disclose information about systems and their operation.

Investment priorities depend on each country's starting point

Governments face different conditions, from electricity access and internet reliability to the strength of schools and public institutions. A practical response begins with reliable power, affordable connections and devices, then builds skills, computing access, representative data and credible safeguards.

Advertisement

The priorities identified in the report cover seven connected areas:

  • Invest in affordable broadband, resilient electricity and community access points so households and small firms can connect reliably.
  • Expand access to computing through regional AI hubs, public cloud credits and research networks, helping universities, startups and public agencies train and adapt models.
  • Scale training for women and young people, combining foundational digital skills with practical learning for health, agriculture, education and small businesses.
  • Collect and govern data representing rural, indigenous and low income communities, and require lenders to check AI credit scoring for discrimination.
  • Support local languages and dialects, open datasets and open source projects that serve public needs.
  • Strengthen privacy protections, algorithmic transparency, assessments and independent audits for uses with high risks.
  • Protect workers through social support, job matching, wage insurance and incentives for employers to create new roles.

Regional differences show why investment must match local needs. In Europe, Denmark, Germany and Switzerland score well on digital infrastructure, skills and governance. Eastern European countries including Albania and Bosnia and Herzegovina lag behind their western counterparts on many of these measures.

Singapore has combined high quality connectivity with public sector innovation and clear guidance for responsible AI. South Korea has combined industrial policy with skills programs and investment in computing capacity. These examples show how policy choices can connect technological development with stronger public institutions and a better prepared workforce.

UNDP's central message is that technology needs investment in people, institutions and fair competition to deliver benefits broadly. National measures and international cooperation can work together to reduce gaps in access and readiness, rather than allowing today's differences to grow as adoption accelerates.

Key Points

  • UNDP warns that AI could widen inequality between countries after five decades of convergence.
  • Asia Pacific has more than half of global AI users, but only about 14 percent of its people actively use AI tools.
  • AI could add around 2 percentage points to annual regional GDP growth, while ASEAN could gain nearly $1 trillion over the next decade with broad and inclusive adoption.
  • Women and young workers face particular employment pressures, and 1.3 billion workers in the region are in informal employment.
  • Unrepresentative data can restrict access to credit, while imported models may struggle with local languages and contexts.
  • Public service examples in Bangkok, Singapore and Beijing show practical benefits alongside infrastructure, security and governance challenges.
  • About 100 companies account for 40 percent of private AI research investment, while 118 countries are absent from international governance discussions.
  • Priority responses include connectivity, skills, shared computing resources, representative data, local language tools, enforceable safeguards and worker protection.
Share This Article

You May also Like