UNICEF and Arm Launch AI Initiative Focused on Children and Emerging Economies

By Suad Seferi · Aug 25, 2026

UNICEF and chip designer Arm are expanding a global programme aimed at bringing practical artificial intelligence tools and AI education to children and communities that are often left out of the technology's fastest-moving markets. The initiative, called AI in Play, brings together technology companies, local innovators, governments and young people to develop AI projects focused on education, health and climate resilience. UNICEF formally named Arm as a founding partner of the initiative on August 20, with Arm detailing the programme in a separate announcement on August 24. The goal is not simply to provide access to existing chatbots. AI in Play is designed around local projects, including tools that can work on affordable devices and in places where internet connectivity is unreliable. By 2030, UNICEF says the programme aims to reach at least 2.5 million children across 12 countries, support at least 20 AI-powered ventures and provide foundational AI literacy training to 200,000 children aged 10 to 16. Half of those learners are expected to be girls. AI that works without the cloud One of the more practical parts of the initiative is its focus on what is known as edge AI. Instead of sending every request to a remote data centre, some AI systems can operate directly on a phone, computer or small device. That matters in communities where fast internet access is expensive, unreliable or simply unavailable. UNICEF is already supporting projects built around that idea. In Laos, AI-enabled devices are being used for air-quality monitoring. In Colombia, a virtual reading assistant called Bookbot helps children practise reading. Other projects are adapting language models for languages spoken in West and Central Africa. The language work is particularly relevant in regions where major AI systems often perform poorly because there is far less training data available than there is for English and other widely digitised languages. UNICEF says work with local organisations in West Africa has already improved speech-recognition performance for Bambara, one of the region's major languages. The underlying idea is straightforward: AI is considerably less useful if it cannot understand the language people actually speak or if it stops working as soon as an internet connection disappears. Children as users, but also builders AI in Play also puts emphasis on teaching children how technology works rather than treating them only as consumers of it. One of the programmes already operating within the initiative is Tinkering with Tech, which introduces children to coding, problem-solving and basic AI concepts through practical activities. The programme has already been introduced in Honduras, the Maldives, Vietnam and Montenegro, with expansion underway in other countries. In Montenegro, UNICEF says around 16% of children aged nine to ten are now coding, building and solving problems with micro during a single academic year through the programme. That makes AI in Play more than a distant global development for the Balkans. One of its educational models is already operating in the region. The programme also works with teachers, not only students, and is designed to adapt to the infrastructure available locally. In some countries that means Raspberry Pi computers or micro. In others, activities may be designed to work with very limited hardware. Who gets to build AI? There is a larger issue behind the initiative. Most of the world's leading AI systems are currently built by a relatively small number of companies concentrated in a handful of countries. That affects which languages receive support, which problems are prioritised and which communities have the technical resources to develop their own tools. AI in Play is trying to widen that group. UNICEF says it wants 45% of the AI ventures supported through the programme to have female founders, while at least a quarter should involve innovators under the age of 30. Arm's role includes providing technical knowledge around hardware and edge computing, while UNICEF contributes its existing relationships with schools, governments and local innovation communities. Other technology companies are being invited to contribute tools, infrastructure, engineering support, mentoring or investment. That means many of the programme's targets are still ambitions rather than completed results. The 2.5 million children, 20 ventures and 200,000 AI-literate young people are goals for 2030, not current achievements. What exists today is the network of programmes already being developed underneath that ambition, including projects in education, language technology, health and climate monitoring. For the Balkans, Montenegro offers an early example of what that can look like at classroom level. The interesting part may not be giving children another AI product. It may be giving them enough understanding of the technology to eventually build something useful themselves.