ITU's GENIE.AI gives governments an open-source toolkit for public-sector AI

A framework built for limited resources

Cosmas Luckyson Zavazava, Director of ITU's Telecommunication Development Bureau, writes that many public institutions face significant barriers before they can even begin adopting AI. He lists limited resources, technical capacity and infrastructure, as well as a lack of access to expertise.

ITU began devising the framework in 2024. A series of consultations shaped a high-level architecture for a generic reference implementation, which was adopted in January 2025, followed by technical development. The project was code-named GENIE.AI. According to ITU, it is designed to fit each country's priorities while strengthening local capacities, technological independence, and influence over countries' digital futures.

ITU says the platform answers a call from developing countries, which requested ITU support in strengthening AI infrastructure, readiness and capacities in a resolution on AI technologies in telecommunication development adopted at the World Telecommunication Development Conference (WTDC-25).

Pilots and country projects

The Global GenAI for Good Challenge, launched last year with IEEE (Institute of Electrical and Electronics Engineers), put the approach into practice. AI developers worked on real public-sector use cases in Bangladesh, The Gambia, and Lesotho. Three winning teams were selected for one-year pilot projects, which ITU says are now approaching completion.

GENIE.AI also builds on ITU's Open Source Ecosystem Enabler (OSEE) programme, in place since 2023, which helps countries develop skills, institutions and local tech ecosystems. Two examples are named:

• Kenya: ITU has equipped government agencies and the local developer community with expertise in open-source AI, with a functional service set for release within months.
• El Salvador: the platform provided the foundation for AgroGenio, an AI-powered agricultural advisory prototype drawing on more than 150 national and international agricultural guidelines and manuals. It was developed under the MERIAN project and, according to ITU, could provide a foundation for potential national-scale deployment.

A network of partners

ITU stresses that the growing network around GENIE.AI matters as much as the technology. Countries contribute real-world challenges and local knowledge. UN agencies including the Food and Agriculture Organization (FAO), the World Health Organization (WHO) and the World Meteorological Organization (WMO) bring domain expertise. IEEE connects the initiative with AI innovators and solution developers, while corporate technology partners such as Intel and IBM contribute open-source technologies and expertise.

According to ITU, the involvement of local technical teams means the resulting solutions can be adapted, maintained and sustained.

What it means for countries

ITU's longer-term vision is local implementation of GENIE.AI combined with international knowledge-sharing, so that countries can exchange insights and replicate valuable innovations. The organisation describes the platform as its mechanism for moving from AI experimentation towards practical public-sector applications.

For governments with limited budgets and technical staff, the pitch is an open, model-agnostic base they can adapt themselves rather than a fixed product. The article does not give details on costs, rollout schedules for further countries or performance results, so the real-world impact will depend on how the current pilots and the Kenyan service turn out.