Introduction

AI tools are increasingly becoming part of everyday life in Africa. However, most of the AI systems adopted are trained on Chinese, European, and English databases rather than on local languages.  This issue, combined with weak AI regulations, excludes many people from effectively using AI tools. The African Development Bank, UNDP and private partners have launched the AI 10 billion initiative during the Nairobi AI Forum 2026. These resources will fund the adoption of AI through sustainable infrastructure, partnerships and investment plans. While this initiative presents a great opportunity to support inclusive growth across the continent and create more jobs, it is key that local, specific development plans follow the announcement. Failure to do so risks perpetuating language exclusion and inequality.

Background

The AI Forum was co-hosted by the Governments of Kenya, Italy, the UNDP and private sector executives in Nairobi and its objective was to discuss how AI can be sustainably adopted in in various sectors including healthcare, education, farming and banking and what infrastructures are needed to foster inclusive progress (UNDP, 2026) across the continent. Among other initiatives that were launched, the AI 10 billion initiative, is a partnership between the African Development Bank, UNDP, AI Hub and private partners to invest in digital opportunities with the aim of creating 40 million new jobs in Africa by 2035. The initiative will fund AI entrepreneurship, data infrastructures, and technical support for AI policies and hubs, foster computing capacity and train engineers and analysts. In addition, the initiative aims to establish education programmes to train local populations in the use of AI with the aim of enabling a safe digital transformation at the continent’s scale (Taki, 2026).  

Implications

The project aims to foster collaboration between governments and local realities. If implemented strategically, it has the potential to be a defining step towards positioning Africa as a leading contributor in the AI area and creating a new labour market. However, if not planned correctly and in partnership with local knowledge, this project could have the opposite outcome, risking deepening language exclusion and fortifying technological dependency on international actors.  

Firstly, as Yucer (2025) explains, many AI systems are currently trained on English, Chinese, or European datasets (iAfrica, 2025). Despite English and French being the official languages of 23 and 21 countries, respectively, an additional 2,000 indigenous languages are spoken on the continent (Chepkemoi, Kipchirchir, 2025). However, AI systems do not currently cover all these languages yet. As oral traditions dominate, many African languages are not yet digitalised, and as such, there is a lack of data to train LLMs. In addition, many AI tools were built to use Latin scripts, so training them on non-Latin data is more expensive and risks importing biases from English, misrepresenting local contexts, and collapsing language and cultural diversity (Lynch, 2025). It follows that AI tools may currently not benefit most Africans. As a result, expanding the use of AI without analysing which kind of AI is being used could further exclude people from jobs rather than foster inclusion. Indeed, people who do not speak English or French will be disadvantaged as they cannot access the more advanced AI systems. For instance, the AI initiative aims to fund projects that utilise AI tools to support health workers, provide medical information, and improve diagnostic accuracy, thus reducing the burden on healthcare workers (AfDB, 2025). However, Reed (2024) explains that there are limited clinical datasets for training purposes in minor languages, meaning that health information is not yet available in many of them, and many people risk missing out on healthcare. Similarly, AI tools can process information and enable farmers to access climate, soil and market information. However, data is more abundant in English and French, so those who do not speak these languages cannot access the latest information (AfDB, 2025). 

Secondly, there are currently few regulatory environments across the continent, and legislation on data use and privacy is poor, which does not compel developers to invest in the inclusion of African Languages in AI systems. In 2024, the African Union Executive Council endorsed the Continental AI Strategy in line with the 2063 Agenda and the Sustainable Development Goals (SDGs) (African Union, 2024), encouraging its members to implement AI governance and regulations. In practice, AI regulation remains fragmented as different countries are at different stages of the regulatory process. For instance, Kenya and South Africa's data protection laws are influenced by GDPR, and Ghana is using UNESCO’s Readiness Assessment Measurement (Regulations.ai, 2026). Other countries, such as Somalia and South Sudan, have yet to begin developing AI strategies and policies. (Research ICT Africa, 2025) Currently, there is no continental strategy for training AI models to be representative of Indigenous languages and cultures, and there is no common regulatory framework that encourages linguistic inclusivity. This is combined to the fact that projects such as Meta’s No Language Left behind (NLLB) offer models that can be customised for African languages, follow international rather than local regulations and raise concern about data transparency, specifically around how linguistic data extracted from communities is used (Cooper, 2026) Without a clear governance framework, investing in these projects, could strengthen the dependence on corporations and undermine the development of national regulatory frameworks (Nsubuga, 2025), enforce the exclusion already embedded in AI systems.

Recommendations 

The Nairobi AI Forum underlined the need for databases, ethical AI governance and local workforce upskilling (ADBG). It is key that international funding does not finance policy briefings alone but also collaborates with local initiatives and research hubs that have already proposed solutions for an effective and inclusive implementation of AI tools.

1.      Support and scale community-driven initiatives such as African Next Voices, Masakhane, Lelapa AI and LinguAfrica (iAfrica, 2025) that collect and digitize local language data, ensuring that AI development reflects linguistic diversity.

2.      Fund the development of natural language processing tools tailored to African languages, including non-Latin scripts.(Ibid.). 

3.      Expand access to computational resources and training programmes to enable local researchers and developers to build context-specific AI systems, so that a larger portion of the population can benefit from jobs being created by AI. 

4.      Foster ethical AI governance and collaboration between governments, private sectors and local builders of infrastructures and support the creation of AI regulations, in line with the Continental AI strategy launched in 2024.  

5.      Empower local startups through projects such as the Harmonic Africa Startup Acceleration Programme, which equips African AI startups with capital and technical assistance to research on feasible solutions in agriculture, health, education and energy (UNDP,2026).

Conclusion 

While the 10 billion initiative is a great opportunity to foster digital growth in Africa and promote international collaboration, this announcement must be followed by a detailed, inclusive, and sustainable plan to accelerate digital transformation on the continent. Currently, the most advanced AI systems utilise French and English, leaving a large sector of the population excluded from using AI. However, many local initiatives are underway to both train AI systems in local languages and advance AI regulations. Thus, if planned strategically, the AI 10 billion investment plan can be a great opportunity to address the implications of language bias in AI model training and poor ethical governance.