Economy10 October 20264 min read
IT Ministry launches World Bank report urging developing nations to adopt and adapt AI
The Ministry of Electronics and Information Technology launched the World Bank Group's World Development Report 2026, 'The Promise of Artificial Intelligence', at an event organised under the IndiaAI Mission with the World Bank Group. The report was first released globally in August 2026.
The report argues that AI can drive development only when supported by infrastructure, skills, data and strong institutions. It says poorer countries gain more from adopting and adapting AI than from trying to build frontier models.
What the report finds
AI is spreading faster than earlier general-purpose technologies. Steam took 80 years, electricity 40 and the internet 20 to reach lower-income economies, while ChatGPT drew 50% of its traffic from middle-income nations within six months. Use is still uneven. By April 2025, ChatGPT was used by 25% of internet users in high-income economies, against 5.8% in upper-middle-income, 4.7% in lower-middle-income and 0.7% in low-income economies. Historic gains, the report says, have gone not necessarily to the inventors of such technologies but to countries that quickly build the complements and deploy applications at scale.
It frames AI's development value through the Three Cs: Capabilities, Concentration, Complements. On capabilities, AI can automate non-routine cognitive tasks such as disease diagnosis, weather forecasting and farm advice, easing acute shortages of specialists. On concentration, advanced compute, chip design and frontier models sit mainly in the US and China, and top US technology firms are projected to spend over USD 750 billion on AI infrastructure in 2026. This creates dependency risks while offering cheaper access to pre-trained models. On complements, AI needs reliable physical and institutional foundations. In Sub-Saharan Africa, 32% of rural schools lack electricity and 68% lack internet, and 89% of 10-year-olds struggle with foundational reading (2024).
The report points to cases where AI has delivered. Wadhwani AI's Cough Against TB in India reached 90% screening accuracy. The Rori learning platform in Ghana delivered a year of learning gains at about $5 per student. AI weather forecasts saved as much as USD 560 per small farmer in Telangana. Kenya used AI to allocate more than 10,000 court disputes, and daily diabetic eye screenings in Bangladesh rose 40%. It also notes that misaligned tools, such as failed automated cash transfers, expose persistent data risks.
Risks the report flags
Generative AI threatens to automate 4.5% of jobs in developing countries against 14.2% in high-income economies, but productivity gains remain skewed towards advanced nations. Business process outsourcing (BPO) gigs fell 39% in 2025, and domestic data centres increasingly strain local power and water grids.
Kenyan data labellers earn 1.32 to 2.00 an hour. Africa has 66 fact-checking bodies for 1.5 billion people. Over 80 countries had released national AI strategies by mid-2026, yet Rwanda is the only one of 25 low-income countries with an official AI policy framework. The report says full domestic AI sovereignty is cost-prohibitive for developing nations, which therefore need open-source models and regional compute sharing.
What the report recommends
The report proposes an 'Adopt, Adapt, Advance' ladder. Countries should start with off-the-shelf open-source and commercial tools for immediate bottlenecks such as administrative triage and basic translation, though adoption alone is necessary but insufficient. The largest dividends, it says, come from adaptation: tailoring models to local languages, agro-climatic conditions and domestic healthcare protocols, and preventing misdiagnoses caused by Western-trained data. It also asks for multimodal tools, such as voice-based applications over basic 2G or feature phones. On advancing, it advises against building frontier foundation models from scratch, which need billions in compute and massive energy grids, and favours Small AI, meaning compact, low-compute models for specific uses.
Governments are given three roles. As enablers, they should provide reliable electricity, affordable broadband, digital literacy and foundational education, and widen access to compute and quality public datasets through Compute-as-a-Service and open data repositories. Labour and tax policies should prevent premature job displacement.
As users, they should move from fragmented pilots to evaluated, scalable solutions, using public procurement to test and scale domestic applications in agriculture, healthcare, education and social protection. Standardised, clean, AI-ready government data integrated with Digital Public Infrastructure (DPI) can improve service delivery.
As regulators, they should adopt agile, risk-based rules, beginning with voluntary standards, algorithmic impact assessments and third-party audits. Existing consumer protection, competition, data protection and cyber laws should be enforced against AI-enabled fraud, deepfakes and algorithmic discrimination, and international coordination can reduce fragmentation. The report also urges regional compute clusters, fair wages and labour protections for data-annotation and content-moderation workers, and stronger media literacy, fact-checking and synthetic-media watermarking against disinformation.