Masakhane: The Story of Building Together

By Sakhile Dube 

PART 2 - Who gets to build Africa’s AI?

Building without extracting

At the heart of Masakhane is the African philosophy of Umuntu Ngumuntu Ngabantu, meaning a person is a person through other people.

It shapes not only the community's culture but also its research approach.

One of Masakhane's deliberate positions is its rejection of what it calls parachute research, when outsiders drop in, collect data, and leave, without engaging or benefiting the local community.

That question of who builds the technology becomes even more important as global interest in African-language data grows. 

For Masakhane, the people generating data cannot simply be treated as sources or hired annotators. They should be recognised as researchers, contributors and owners of the knowledge their communities produce.

Jade Abbott, Masakhane co-founder, is critical of research institutions and technology companies that collect African language data without meaningfully involving the communities that produce it, or ensuring that those communities benefit from the technologies built from it. 

Masakhane's goal is not simply to incorporate African languages into existing AI systems. It is to reshape who builds those systems, who owns the resources behind them and who benefits from the technology they create. 

“African-language technology should not be something developed about African communities or for African markets. The communities themselves should have a meaningful role in determining what gets built, how the data is governed and where the resulting value goes,” says Hatem Haddad, Masakhane Research Foundation board member (at the time of the interview).

From a community to an ecosystem

In 2021, Masakhane earned international recognition, including winning the Wikimedia Media Foundation award. The honour affirmed both the impact of the community’s research and the value of its community-led approach. 

The scale of the work grew.

The major turning point was in 2022, when Masakhane registered as an organisation.

“Between 2019 and 2022, a lot of our work was volunteer-driven, so we had not yet begun receiving funding. By 2022, however, funding opportunities emerged, and we began working through fiscal sponsors,” notes Kathleen Siminyu, Board chair, Masakhane Research Foundation.

Masakhane’s research now spans more than 50 African languages across over  35 African countries. The community has produced more than 40 publications at major NLP venues and hosted more than 10 workshops, co-located with major conferences. Open datasets built by more than 40 GitHub contributors are used in academic research and by early-stage startups.

The community itself has changed.

What began as an informal collective has evolved into the Masakhane Research Foundation, guided by a 2025–2029 strategy for its next phase of growth.

But the most important measure of progress may not be the number of papers or datasets.

It is the change in the question being asked.

The conversation is no longer simply about the absence of data for African languages; there is now an ecosystem building it.

“Success has already happened for Masakhane. If you look at the number of publications coming out of us, Masakhane wins, and I'm not even involved in any of them anymore”, says Abbott.

The stakes have changed

Given Masakhane’s origins, the African language ecosystem is becoming increasingly important as AI expands beyond research labs into everyday infrastructure.

Language inclusion is no longer solely an NLP research challenge;  it is becoming a matter  of access, safety and governance.

As AI systems are deployed in education, healthcare, agriculture and public services, the languages people understand, and the contexts they can interpret, will increasingly determine who can use them effectively. 

Haddad says this makes linguistic exclusion a governance problem. 

“If policymakers are deciding how AI should be developed and deployed without accounting for the languages spoken by the people affected by those systems, entire communities risk being left out of decisions about technologies that will shape their lives,” he says.

For Masakhane, the answer is not to wait for global AI companies to solve the problem. Instead, it is to continue building the datasets, models, research capacity and institutions that enable African communities to shape how they are represented in the AI systems that will increasingly shape their lives.

Meanwhile, foundational models are scaling rapidly.