As AI rapidly reshapes global economies and public services, a profound silence persists at the heart of the digital revolution. African languages, spoken by over one billion people, remain largely absent from the data powering today’s leading AI systems. There are over 2000 African Languages, yet less than one percent of widely used training data reflects African linguistic contexts. This creates structural barriers that risk entrenching digital inequalities, excluding already marginalised groups from accessing social, political and economic rights and erasing cultural heritage. To address this, we must reform data governance within the United Nations Global AI Governance Dialogue by platforming value systems that focus on community, Ubuntu, human dignity, and indigenous knowledge.
The foundation: People at the centre of the governance framework
At the Masakhane African Languages Hub (the Hub), governance is not merely a set of technical rules but a reflection of the philosophy of Ubuntu: “I am because we are”, which ultimately centres people in our differences, social contexts and nuanced lived experiences. Within AI discourse, this means recognising that data is not an abstract commodity to be harvested; it is a living expression of community identity.
Ubuntu reminds us that meaningful change begins by treating African populations as co-creators rather than passive data subjects. This requires a shift toward community-centred governance, where decision-making processes prioritise multi-disciplinary approaches – including local linguists, culturalists, tech builders, sociologists, and civil society speakers – coming together. To ensure human dignity, projects must actively involve community members in design and execution, ensuring all contributors are treated with respect in spaces free from exploitation.
From extraction to sovereignty
Historical power imbalances in the global Natural Language Processing (NLP) ecosystem have traditionally led to external control over research agendas and to extractive data practices. To address this, the Hub advocates for African linguistic sovereignty – the ability of African languages to thrive in the digital sphere with local ownership and control. This involves two critical shifts:
- Non-extractive collaboration: Partnerships must move beyond the funder-recipient model. Through our positioning as a standard-setter, we ensure that value generated – whether through data, intellectual property, or research – primarily benefits and remains accessible to the African ecosystem.
- Technical sovereignty: True agency requires building on independent African infrastructure. Reform must prioritise investing in building fit-for-use infrastructure that allows the creation of local language models (small or large) that work in the context of limited connectivity, power challenges and affordability of infrastructure.
Ethical stewardship and purpose limitation
Governance reforms need to move beyond “check-box” consent toward ethical stewardship. This involves the implementation of robust protocols rooted in transparency and accountability:
Full transparency: Participants must receive comprehensive information about consent and data usage in a language and format they easily understand.
Purpose limitation: Collected data must be used strictly for the purposes for which consent was obtained.
Equitable benefit-sharing: Communities should directly benefit through fair compensation, access to the technologies created, and a clear say in how resulting assets are used.
Standard-setting for global impact
The Hub does not merely participate in the Global AI Governance Dialogue; to the political world, we aim to be the standard-setter. We set the benchmarks that enable intergovernmental bodies to protect their citizens’ linguistic rights. Our “Benchmarking in the Wild” initiative shifts evaluation from abstract technical metrics to real-world conditions, such as code-switching and dialect variation.
We translate this research into high-impact delivery across four priority sectors: healthcare, education, agriculture, and government services. The following active workstreams exemplify this transition:
LINGUA Africa: Connects open language resources to real-world use cases, such as symptom triage systems in health or literacy tools in education.
Project Echo: Develops gender-responsive AI innovations in livelihoods and low-risk health to benefit 20,000 African women.
A roadmap to 2030: Digital agency for the one billion
Our vision is to empower one billion Africans by 2030 with AI technologies in 50+ languages that reflect their linguistic diversity and cultural contexts. This is an act of digital agency and dignity.
As a strategic architect, the Hub influences global funders to diversify models toward long-term technical sovereignty. We work towards ensuring every language has a clear fit-for-purpose pathway to technical autonomy, regardless of commercial interest.
Every time we speak with clarity in these global dialogues, we decentralise power. By architecting a sovereign future, we ensure the AI revolution is not an era of linguistic erasure, but a catalyst for cultural preservation and socio-economic resilience across the continent.
