Picture this - a team of researchers from a foreign country arrives in one of the low-income settlements in Nairobi.

Through the local community leadership, they undertake an exercise to measure blood pressure, check blood sugar, photograph eyes with a retinal scanner for hundreds of residents.

 Two years later, an  artificial intelligence (AI) tool trained on these data is shown to predict heart disease and stroke risk up to a decade in advance.

 The AI trained tool goes on to fetch hundreds of millions of dollars in user fees once deployed by the owners.

In return the people who took part in the research get nothing from these proceeds.  This situation is increasingly becoming a reality in Kenya.

Just recently, a friend in the medical technology industry was approached by a European inventor of such a tool, this time to help them perform a – proof of concept study - in a Kenyan slum since the regulatory requirements in Europe proved too demanding for them.

Needless to say, the request was rejected after back and forth and discovery of the intention behind the request.

AI-enabled medical devices trained on health data from low-income populations are already being developed and sold globally. The more diverse the data, the smarter the devices could become – supposedly.

In the above case of an AI-enabled medical device trained on hundreds of Kenyan patients — carrying unique mix of genetic heritage, diet, tropical disease exposure, and urban poverty stress — likely producing a unique disease prediction model that cannot be replicated from a western country dataset alone, making such tools commercially precious.

A 2025 study by Occhipinti on the recessionary pressures of generative AI published in the Bulletin of the World Health Organisation, warned that without proper interventions, AI will create a world where a small group of technology owners grow extraordinarily wealthy while communities and developing nations fall further behind.

The researchers showed that economic exclusion directly causes mental illness, substance abuse, and what they termed as “deaths of despair”; — people dying not from disease but from hopelessness.

While most African countries are likely to face this reality, Kenya is perhaps better positioned thanks to three policy developments that have converged in the recent past.

First, in 2025 the Pharmacy and Poisons Board launched guidelines to regulate AI-powered medical tools, including Software as a Medical Device. A significant step — but the guidelines focus on safety standards, not on who owns the training data or who profits when these tools go global.

Second, Kenya launched its National Artificial Intelligence Strategy 2025–2030, naming data sovereignty as a national priority. 

The strategy references the Data Protection Act of 2019 and calls for agile regulatory frameworks. It acknowledges Kenya must control its own data but seems to stop at that without providing actual tools. 

Third, the Artificial Intelligence Bill 2026, sponsored by Senator Karen Nyamu, has been tabled in the Senate. It proposes among others mandatory oversight of high-risk AI. It follows a High Court ruling from February 2026 demanding urgent AI regulation.

While these interventions are great, none is yet to substantively answer the central question about what communities get when their data trains a tool that makes someone else rich?

Therefore, it is time Kenya put in place a practical commercialisation mechanism and framework that protects, compensates citizens, and creates a lasting claim on the AI systems built from our individual and aggregated data.

Four interventions could make this a near reality. First, putting in place informed community consent that ensures that no data collection from Kenyan communities for AI training without plain-language consent approved by a

community representative body.

Secondly, a mandatory data provenance registration that ensures that any AI developer using Kenyan data registers the dataset with the Data Protection Commissioner and attach a unique digital identifier that follows the data into every product it trains.

Thirdly, a data dividend scheme in form of a legislated royalty that ensures that when an AI product trained on Kenyan data generates commercial revenue above a set threshold, a percentage flows into a Kenya Data Sovereign Fund, distributed to originating communities, with a portion funding other public good services such as health and education.

Finally, the Office of the Data Protection Commissioner should invoke the establishment of community data cooperatives that negotiate collectively on behalf of Kenyans — moving beyond individual consent forms.

In Kenya, data is the new land...and we know what happened the last time Africans did not own their land.

Authored by Dr. John M. Mwangi, Health Policy Expert based in Nairobi