Artificial intelligence (AI) language models must be trained to meet Africa’s diverse story, says communications minister Solly Malatsi.
Malatsi was speaking at a ministerial forum that took place ahead of Africa Tech Festival 2025 in Cape Town. The annual forum, which brings together ICT ministers and dignitaries, is jointly hosted by the communications ministry and African Telecommunications Union (ATU).
Held under the theme: “AI powering Africa’s leap into the digital and intelligent era”, the minister reflected on the requirements to achieve AI goals, such as local language resources.
He also highlighted the need for guiding values and cooperation to scale the things that work. According to Malatsi, Africa’s approach to AI must be rooted in dignity, fairness and accountability.
“If AI is to work for Africa, it must embrace the diversity of African languages, in our contexts and for our use cases. To achieve this, we can’t solely rely on the investment agenda of other countries.
“We must invest in models trained on African data and languages. This is the only way to ensure that people can unlock the potential of AI, regardless of whether they do so in Chichewa, Swahili, isiZulu or English.
“It also means that models must be trained to understand our climate, health burdens, informal economies and diverse cultural contexts. It means that the dominant models understand that Africa’s story is not a singular, but one that is rich and diverse.
“We require both public and private investment in local computing capacities, regional cloud centres and energy systems that make model training and inference viable in our countries. We must also empower our people with the skills required to benefit from the potential of AI, without them needing to become experts in AI.”
As AI development advances,some concerns have been expressed about AI’s power to deepen inequality between nations instead of narrowing it.
Resultantly, calls have been amplified to dismantle some of the barriers that would exclude Africa’s one billion population, with language a key priority.
Given the continent’s multilingual landscape, it is feared AI’s rapid advancement will have the potential to leave much of Africa’s 1.2 billion population behind.
For example, West African nation Nigeria has more than 500 language dialects, while SA has 13 official languages, but only one in 10 South Africans speak English at home – the language that dominates the internet. Meanwhile, common languages spoken throughout the rest of the continent include Arabic, French Creole, Shona, Swahili and Swati.
For Malatsi, achieving these AI principles will require pragmatic steps.
African models must be built for African needs. He said this can be achieved through research collaboration and investing in access to representative datasets, with clear protocols for consent, privacy and community oversight.
“We must also back African researchers and start-ups, and ensure they have access to the necessary compute resources.”
Equally, there must be promotion of cultural knowledge and it must be protected, to ensure fair compensation. “The depth of our cultural heritage is not a free input. When cultural expressions, traditional knowledge, or community archives are used to train models, it must be lawful, respectful, and there must be fair compensation.
“This is not only a matter of justice, but a matter of sustainability. Creators and communities will contribute when they are partners, not sources to be mined. We should design licensing schemes, benefit sharing mechanisms, and custodial institutions that make it easy to do the right thing and hard to do the wrong thing.”
ATU secretary-general John Omo stated: “With sound governance, Africa can lead the future of AI. We must strengthen our ecosystems that are necessary for this growth. Let us make Africa a place where its brightest talent can build the future without leaving the continent. The journey forward starts with our collaboration with partners…and other industry players.”
Share