Every Thursday, Delve Into AI will provide nuanced insights on how the continent’s AI trajectory is shaping up. In this column, we examine how AI influences culture, policy, businesses, and vice versa. Read to get smarter about the people, projects, and questions shaping Africa’s AI future. Let us know your thoughts on the column through this form.
This weekend, on the sidelines of the United Nations General Assembly in New York, Nigeria’s Minister of Communication, Innovation and Digital Economy, Dr Bosun Tijani, unveiled N-ATLAS, an open-source language model that can recognise and transcribe spoken words and generate text, in Yoruba, Hausa, Igbo, and Nigerian-accented English, joining a growing field of AI tools trained to handle how Nigerians speak and write.
For a ministry that has sought to position itself as a leader in “responsible and inclusive AI,” the timing and venue of the announcement helped the ministry reinforce that positioning. N-ATLAS was built in collaboration with Awarri, a Lagos-based startup, and reflects months of work since the government first announced the project in April 2024. . The models are currently free to access and use for research, prototyping, and smaller-scale applications. However, once they are deployed in commercial settings with more than 1,000 end-users, a separate licence is required.
By making these local language tools more available to Nigeria’s AI ecosystem, the project positions Nigeria as a participant in the global race for “AI sovereignty.” However, its real significance lies in whether local researchers, developers, startups, and policymakers can turn this symbolic launch into practical tools for classrooms, clinics, and farms, despite persistent funding and infrastructure gaps.
Foundational tool for the ecosystem
At first glance, the average Nigerian hoping to chat with N-ATLAS will be disappointed. The model is not yet available as a consumer-facing app. Instead, it lives on Hugging Face, the open-source repository popular with developers and researchers.
“The first release now is literally geared towards developers,” says Sunday Afariogun, lead project engineer at Awarri. “An average user might see the news and think, ‘this is great,’ but they don’t really know what they’re going to do with that information.”
N-ATLAS is more of a foundational layer than a finished product. By offering a base model that understands Nigerian languages and accents, Awarri hopes that others in the ecosystem will build sector-specific applications that reach everyday users in fields like healthcare, agriculture, and education.
For developers, the gap N-ATLAS could fill is significant. “Most of today’s widely used AI models are trained on data that doesn’t fully represent our languages, contexts, or cultural nuances,” says Bilesanmi Faruk, co-founder and CTO of Lena, a Lagos-based edtech firm. “That means we spend extra time adapting global models, or worse, we end up with tools that don’t work well for local users.”
Faruk has already begun experimenting. “I’ve hooked the Nigerian-accented English and Yoruba speech recognition model up to a microphone for live translation,” he explains. “We’re building it into our offline app at Lena so that kids can learn in their native language and still get feedback. That way, we deliver world-class learning in rural areas while maintaining cultural context. It opens up the next 100 million learners.”
Zainab Tairu, a natural language processing (NLP) engineer and researcher, sees similar promise. She has been working on a machine learning project on medication management and struggled to find reliable datasets. “Getting access to a local, open-source model was a big hurdle,” she says. “Having something like this makes it easier for researchers like us, and many others, to build solutions with African voices and contexts at the core.”
Although still invisible to most citizens, the models could play a role in the ecosystem’s progress. Developers are already imagining sector-specific applications. In theory, this could mean farmers calling a hotline and asking about crop diseases in Igbo, patients could describe symptoms in English accents that AI tools actually understand, and receive relevant, contextual responses.
Joshua Firima, co-founder of KrosAI, a voice and text infrastructure AI startup, believes “phone-based AI systems that reach citizens directly” are the next step. “Success will be when Nigerians use AI daily in their own languages without even thinking about it,” he says.
To get these specialised applications, Faruk believes more needs to be done to bring the developers and researchers on board. “Documentation and community support also matter; without them, even open-source models can remain underutilised.”
Lowering barriers, but not eliminating them
While N-ATLAS signals possibilities for the broader AI ecosystem, it also highlights what still constrains the country’s AI space: data, compute, and funding.
“We cannot decide that as a country we’ll wait until we have infrastructure before building software and solutions. If we did, we would fall further behind,” says Afariogun. Nigeria’s data centres are improving, but few can host the GPU racks needed to train large AI models.
Currently, Awarri relies on foreign cloud providers, including AWS and Google Cloud, which remain the more reliable option, but issues of national sovereignty over AI systems still pose a challenge.
“If we decide that we’re gonna wait for infrastructure before we start providing solutions. Then we just lag behind further. So we have to try to do this simultaneously,” he explains.
Access to infrastructure remains a significant challenge for developers expected to adopt the N-ATLAS tools in their AI applications. “Training and fine-tuning still require significant GPU power, which is often out of reach for small teams,” says Faruk. “The cost of cloud credits makes it worse.”
Getting reliable data to support AI projects is another bottleneck. While Awarri built LangEasy.ai to collect thousands of voice samples from fellows in the government’s 3 Million Technical Talent (3MTT) programme, most researchers still collect their own domain-specific datasets. “Starting a new project often means you can’t find existing datasets,” Tairu says. “That makes the whole process very tedious, from collation to deciding what a quality benchmark even means in a Nigerian context.”
Funding is another unavoidable constraint. The government has supported 20 peer-reviewed AI research papers through an 18-month program, but for a country of over 200 million, the output is modest. Without more tangible support for compute access or grants for research, the expected applications for N-ATLAS may never get built. If these bottlenecks persist, N-ATLAS risks being a symbolic achievement rather than a catalyst. Egypt, for instance, which hosts Africa’s top-ranked university for AI, has pledged to produce 6,000 AI research outputs by 2030. Nigeria will need to improve its approach to tangible support if it hopes to keep pace.
The open-source nature of N-ATLAS also has limits. Licensing allows experimentation and prototyping at no cost; however, products that scale beyond 1,000 users will require formal arrangements. That is not unusual in the open-source world, but it does mean that without sustained funding, local developers may struggle to commercialise their work.
The road ahead
Where N-ATLAS could have the deepest impact is in fostering shared ownership of Nigeria’s AI future.
Joshua Firima, founder of KrosAI, says initiatives like N-ATLAS open the door to deeper collaboration within the private sector. However, he adds, their impact will depend on how widely they are distributed and whether the right support structures are in place, including integrating AI into everyday channels such as phone lines and WhatsApp, as well as creating incentives like grants and funding for innovators. Equally important, he argues, are safeguards on privacy, bias, and accountability to build public trust.
Cultural preservation adds urgency for N-ATLAS. “If we have over 2000 languages in Africa, and less than 2% are being represented right in AI, what that then means is that a lot of languages will go into extinction in the nearest future,” says Aizehi Itua, Vice President of Marketing and Communications at Awarri.
N-ATLAS’s success will ultimately be measured by whether startups, researchers, and other ecosystem players can utilise it to develop relevant tools that matter in classrooms, local hospitals, and farms.
We would love to know what you think about this column and any other topics related to AI in Africa that you want us to explore! Fill out the form here.
Mark your calendars! Moonshot by is back in Lagos on October 15–16! Meet and learn from Africa’s top founders, creatives & tech leaders for 2 days of keynotes, mixers & future-forward ideas. Get your tickets now: moonshot..com
