ITU Launches AI Lab For Developing Nations

The International Telecommunication Union has confirmed the AI for Good Lab, launched on 8 July at the AI for Good Global Summit in Geneva, as a global initiative to help developing and emerging economies build the skills, infrastructure and policy foundations artificial intelligence requires — including local datasets, locally adapted models, and in-country compute capacity.

It builds on an approach the ITU has already been testing through its AI for Good Sandbox programme, which has run pilots in ten countries, Cameroon, India, Mozambique, Nepal, Peru, Tanzania, the United Arab Emirates, Uzbekistan, Zambia and Zimbabwe, working alongside the African Telecommunications Union as a regional partner.

What the Lab will do

The Lab is structured around three connected pillars, meant to support countries wherever they are in their AI journey. The first tackles AI readiness and policy, helping governments run capability assessments, spot gaps, and shape national adoption strategies through readiness assessments, policy guidance and workshops. The second builds a talent pipeline, AI Skills Development, drawing on the ITU’s existing AI Skills Coalition, its Innovation Factory accelerator programmes, robotics initiatives, hackathons and machine-learning challenges aimed at youth, professionals, entrepreneurs and public-sector leaders alike.

The third  and the one that speaks most directly to the compute debate that’s followed the continent through Geneva this week; is public AI infrastructure: open datasets, compute resources and open-source toolkits that countries can use to experiment, validate solutions and align with ITU standards. In-country projects under this pillar are meant to develop local compute infrastructure and produce locally adapted, open-weight models, meaning their parameters are published that governments, companies, start-ups and universities can reuse at lower cost.

The stated aim is to move countries beyond isolated pilot projects toward sustainable national AI ecosystems, with countries building datasets rooted in their own context for priority sectors like health, agriculture, education and mobility. Delivery will run through national governments and the ITU’s regional offices, which are expected to establish local and regional hubs.

 The Geneva context

This is where the gap really shows. All week, African delegations came to Geneva asking for more than internet pledges and training programmes, they wanted real compute power, trustworthy data systems, and support for AI that works in their own languages. But most of the week’s big announcements, including the $100 billion Partner2Connect milestone and the new AI for Good Global Commission, were about partnerships and promises, not actual machines and infrastructure.

The Lab’s third pillar comes closest to answering that. It points to the exact things African governments have been asking for: data built from their own communities, computing power based in-country, and AI models whose inner workings are shared openly, so they cost less to build on. It also echoes work already underway on the continent, the data sandbox anchoring Rwanda’s new National AI Agency, the local-language dataset work in Ghana’s AI strategy, and the linguistic-diversity gap Kenya’s own AI strategy and homegrown AI startups have been working to close, as global models continue to struggle with local dialects.

But what the ITU hasn’t said yet matters just as much as what it has. There’s no budget attached to this initiative, no word on how much computing power will actually be available, no list of which countries will join next, and no timeline for when local offices will be up and running. The pilot programmes this Lab is built on have mostly been small-scale tests so far, not major infrastructure projects. So whether this becomes something African governments can actually use, or just another plan waiting for funding, will depend entirely on what comes next.

Who it is for

The ITU frames the Lab’s beneficiaries broadly. Researchers and universities get new channels for collaboration. Students and young innovators get support building AI and robotics solutions. Start-ups get backing through the Innovation Factory and accelerator programmes. Governments and regulators get sandbox environments to test solutions in, along with readiness assessments and policy toolkits to shape their national strategies.

For African administrations several of which, from Dodoma to Lusaka and Harare, are already inside the Sandbox pilot, the Lab offers what the ITU calls a clearer institutional pathway to public AI infrastructure. That pathway exists on paper for now. What it becomes will depend on what actually travels down it.