Data and language: why local data decides how well AI serves you

Every AI tool is built from examples. A translation tool learned from millions of translated sentences, a voice tool from thousands of hours of recorded speech, a chat assistant from a large share of the written web. What was in those examples, the tool can do. What was missing, it does badly or not at all.

This explains a great deal about how AI serves people in Africa. Most of the web is in English and a handful of other world languages, so AI tools are fluent in those and stumble in Luganda, Yoruba, Dholuo or Kinyarwanda. Our guide to why AI struggles with African languages explains the gap, including why the same text can cost more to process in some languages, and introduces the African teams closing it: Sunbird AI in Uganda, N-ATLAS in Nigeria, Mozilla Common Voice volunteers in Rwanda, and more.

The same logic applies far beyond language. An AI tool knows little about the prices in your market, the crops in your district, the rules of your regulator or the habits of your customers, because little of that is written down online. Whoever collects that knowledge properly holds something the large tools lack. That is why we keep returning to local data: it is where people in the region can build, not only consume.

"Properly" matters. Data about people comes with rules, and the next section is about them.

You can see this happening already. Ugandan teams have built translation for 31 of the country's languages, thousands of Nigerians have recorded their voices for a national language model, and Kinyarwanda speakers have donated more than 2,000 hours of recorded speech. Each of these is a store of local knowledge that the large tools did not have, built by people who knew what was missing.

Personal data is any information about a person who can be identified, such as a name, a phone number, an ID number, a location or a health detail. All seven countries we cover have a data protection act governing it, with a regulator to enforce it. The organisation that decides why and how personal data is used, the data controller, must follow the law and in most of these countries register with the regulator.

These laws apply the moment you paste a customer's or a parent's details into an AI chatbot, because that is processing their data and usually sending it abroad. Our guide to data protection laws and AI sets out each country's act, who must register, the cases regulators have already brought, and the one habit that keeps you on the right side of all seven: take the person out before you paste.

Data protection laws and AI in Kenya, Uganda, Nigeria and beyond
Seven countries' data protection acts, who must register, and what they mean for your AI use.

Why AI struggles with African languages, and what data changes
Why AI is weaker in your language, and the people building the data to fix it.