UN report warns AI may widen gap between rich and poor

Behind the global excitement around artificial intelligence lies a stark warning from the United Nations Development Programme (UNDP): without deliberate policy choices, AI could widen the gap between the world’s rich and poor.

In its new Human Development Report, UNDP says the benefits of AI are likely to be captured disproportionately by wealthier nations and well-resourced communities unless digital access, education and infrastructure improve for everyone.

The report likens the moment to the “Great Divergence” of the Industrial Revolution, when rapid modernisation in parts of the West left many other regions far behind. Today, AI’s transformative potential risks repeating that pattern unless societies put “people first” rather than “technology first,” said Michael Muthukrishna of the London School of Economics, the report’s lead author.

High hopes, but uneven readiness

A global survey conducted for the report — spanning 21 countries, 36 languages and representing about 63 per cent of the world’s population — found widespread optimism about AI’s potential, especially in developing nations.

Key findings from UNDP’s 2025 AI & Human Development Survey

• Survey covers 21 countries in 36 languages, representing 63% of global population.

• 20% of people have recently used AI in work, health or education.

• 66% expect to use AI in these areas within the next year.

• 64% believe AI will improve productivity; rises to 70% in low/medium-HDI countries.

• 60% foresee new job opportunities from AI; 12% fear job losses.

• 25% of Asia-Pacific still lacks internet access.

These figures suggest significant demand and enthusiasm, but the report warns that the ability to benefit depends heavily on access — to electricity, reliable internet, digital devices, skills training and supportive regulation. Many communities lack these basics, making them effectively “invisible” in AI-driven systems.

Risks extend beyond inequality

Beyond economic divides, the report flags broader risks such as privacy violations, bias in automated systems, deepfakes, surveillance, and cybersecurity threats. Researchers have already found hackers using AI to automate parts of cyberattacks.

Environmental concerns also loom large. Even wealthy countries face challenges as expanding data centres — the backbone of AI — demand huge amounts of electricity and water. Meeting this demand could hinder progress on emissions targets and place pressure on local ecosystems.

Asia-Pacific: sharp contrasts in readiness

The report highlights stark contrasts within the Asia-Pacific region. Economies such as China, Japan, South Korea and Singapore are well positioned to harness AI, with strong infrastructure and investment. By contrast, places like Afghanistan, the Maldives and Myanmar lack the power supply, digital connectivity and human capital needed to take advantage of AI’s potential.

Within countries — even advanced ones — regional inequalities persist. Rural areas and marginalised communities often lag far behind urban zones in connectivity, skills and access.

About a quarter of the Asia-Pacific population still lacks any online access, the report notes, putting millions at risk of exclusion from digital services, payment systems, education tools and job opportunities.

A call for balanced, people-centred AI

Philip Schellekens, UNDP’s chief economist for the Asia-Pacific, said the conversation around AI must be grounded in “more balance, less hysteria and hype.” While AI can generate enormous public value — in agriculture, healthcare, disaster response and governance — effective regulation and broad-based investment are essential.

The report argues that AI is becoming a foundational utility, akin to electricity or the internet, and that governments must expand digital infrastructure, strengthen education systems, ensure fair competition and reinforce social protections.

“The goal,” it concludes, “is to democratise access to AI so that every country and community can benefit while protecting those most at risk from disruption.”

- with inputs from AP