According to the World Bank, artificial intelligence opens a historic opportunity to solve development problems that have been entrenched for decades, but it requires acting strategically and quickly.
Its World Development Report 2026, the first comprehensive assessment of AI's impact on the 6.8 billion people in low- and middle-income countries, places this technology on a par with the steam engine, electricity, computing and the internet.
From 80 years to six months
But this revolution is advancing far faster than any other. Steam took 80 years to reach lower-income countries; electricity, 40, and the internet, 20. With AI, the timelines collapse: barely six months after the launch of ChatGPT, middle-income countries already accounted for half of its global traffic.
Speed opens opportunities, but it does not guarantee success. History shows that those who benefited most from the great technological revolutions were not always those who invented them, but those who built infrastructure, human capital and institutions in time. With AI, the World Bank warns, this is once again decisive.
According to the World Bank, AI's impact on development will depend on three factors: capabilities, concentration and complements. The first is its ability to perform tasks that require specialized knowledge, such as diagnosing diseases, forecasting the weather or advising farmers, especially valuable in countries where these resources are scarce.
The second is concentration. The most advanced models, chips and data centers are in the hands of a few companies and economies. This creates dependency, but it also allows other countries to adapt existing models without bearing the enormous cost of creating them from scratch.
Adopting the AI that already exists
The third factor is complements: infrastructure, education and solid institutions. Without these pillars, AI will struggle to be deployed effectively and adapted to local needs. Faced with the race to build increasingly powerful models, the World Bank proposes a more realistic strategy for developing countries: adopt, adapt and advance. Each step requires more investment, infrastructure and technological capacity.
The last rung, developing their own frontier models, requires advanced chips, large data centers, enormous volumes of information and top-level researchers. Its cost makes it unattainable in the short term for many economies: in 2026, AI infrastructure investment by a handful of companies exceeds the size of the economies of numerous countries.
The alternative is to start by adopting the AI that already exists and applying it to concrete needs, such as medical diagnosis, agriculture or business management. However, it is at the second rung, adapting, where the World Bank identifies especially important potential.
Open models for local AI
Adapting means using open models to create solutions capable of understanding the languages, laws and needs of each society. The report cites Apertus, BERT, Gemma, Llama, Mistral and Qwen as foundations for developing local applications.
Examples already exist such as Latam-GPT in Latin America and SEA-LION in Southeast Asia, designed to respond better to the linguistic, cultural and economic particularities of their regions. The key, therefore, is not to compete to build the world's most powerful AI, but to adapt available technology to solve concrete problems.
Farmers saving up to $560 with AI
In Telangana (India), an AI weather prediction system helped small farmers make better spending decisions, with savings of up to $560 per producer. Kenya, for its part, used AI to distribute more than 10,000 court cases a year among 1,500 mediators, helping to reduce pending cases. Meanwhile, in Ukraine, a Transparency International tool uses this technology to detect suspicious patterns in public procurement and channel citizen complaints.
Among the risks, the World Bank points to algorithmic bias, fraud, disinformation and technological dependence on the few companies, mainly in the United States and China, that control the most advanced systems. AI can reduce development gaps, but it can also widen them. As the most sophisticated models advance, millions of people still lack electricity, internet or basic education.
The state as user and regulator
The contrast is especially stark in sub-Saharan Africa: almost a third of rural schools lack reliable electricity, more than two thirds have no reliable internet connection and nearly nine out of ten ten-year-old children cannot read a simple text. The World Bank therefore insists that the AI revolution must be accompanied by investment in education, electricity, telecommunications and institutions.
The World Bank assigns governments three roles: enabling, using and regulating AI. This means building infrastructure, incorporating the technology into public services and establishing rules that build trust. In countries with weaker institutional capacity, it advises initially avoiding rigid regulations and opting for voluntary standards, impact assessments, independent audits and testing protocols.
The readiness gap is already evident. By mid-2026, more than 80 countries have national AI strategies, but among the 25 low-income countries only Rwanda had one. The paradox is clear: some of the countries that could benefit most from AI are also the least prepared to take advantage of it. The risk is repeating the history of other technological revolutions: generating more productivity and wealth, but distributing its benefits very unevenly.