The Silent Loot of Data in Africa

As of today, Meta has not only refused to pay any of the fines imposed but has also operated a business-as-usual data policy in direct contravention of what it was fined for. Meta has continued to protest privacy demands, considering them unrealistic, even though it operates these same ‘unrealistic’ policies in Global North countries. The company has also faced a similar regulatory standoff in South Africa. Given Africa’s dependence on Meta platforms, there is no competition to compel Meta to adopt complete transparency or data repatriation. 

A CONTINENT-WIDE DATA PROBLEM

This problem is spread across the continent. According to the Open Data Barometer, Africa lags globally, with South Africa, the top-scoring country with 36 points, still significantly below the global average. In the age of AI, a technology built on the backbone of data, this divide widens. Since 2021, tens of thousands of data points have been siphoned from Africa to train AI. By extension, Africans have also been underpaid and overworked to label data that OpenAI used in training its models. Yet, AI models struggle to recognize or replicate African faces and voices. The proposition of data as more than just information, as power, cultural relevance capital and means of control becomes real in the age of AI, and nowhere is this more real than in considering Africa’s AI and data disparity. 

As with countless other resources in Africa, our data is currently extracted, packaged and sold by foreign entities. African data is mined by international tech corporations, non-governmental organizations and foreign governments under the convenient banners of development, security and technological advancement. At the same time, Africa is excluded from the profits and decision-making processes, as well as the governance frameworks that shape its digital future. Since 2021, AI companies have paid out millions of dollars to procure data, either through official agreements or court settlements—none of which has gone to, or is expected to go to, an African entity.

THE GIFT FROM TROY

This discrepancy is symptomatic of the flawed means by which Africa’s digital systems operate. Consider, for example, the growing footprint of international tech giants across African digital infrastructure, whose investments come wrapped as progress: better connectivity, improved access to healthcare, and smarter cities. However, in reality, these initiatives often serve as opportunities to harvest massive quantities of data. They do this in a million tiny ways. When an African user queries the same website as someone from the Global North, for instance, the cookie options are different, with the African user often being forced to accept more tracking tokens than needed. 

Even when data is referred to as ‘open’ or collected for humanitarian reasons, the long-term ownership typically remains with international institutions. African governments, constrained by weak policies, inadequate infrastructures and external political pressure, have yet to establish meaningful control over the digital wealth generated in their nations. This imbalance has left Africa as the supplier of raw materials for the fourth industrial revolution.

The 2024-2025 Cloudscene report shows that out of more than 8500 data centres worldwide, fewer than 90 are in Africa. Most of these are found in South Africa, Nigeria and Kenya. Even when African start-ups seem ‘local’, they are usually registered in other countries, and very few of them keep their data in Africa. 

This system is dysfunctional at every level. African data—not stored in Africa, not taxed in Africa, and not recognized as African-owned—is being used to build Meta’s language model and targeted advertising algorithms. Despite this, the systems built with this data neither understand nor represent African values and cultures. Voice assistants barely comprehend African accents and never speak indigenous languages. 

AI AND THE NEW FACE OF COLONIALISM

Today, colonialism is now less about borders and physical force and more about data and algorithms. In Ethiopian cognitive scientist Abeba Birhane’s essay ‘Algorithmic Colonization of Africa’, she argues that the Global North tech corporations do not impose their will on Africa through armies, but rather by harvesting African data, designing digital trojan horses and compelling economic supremacy. Birhane goes on to write that ‘algorithmic colonialism, driven by profit maximization at any cost, assumes that the human soul, behaviour, and action are raw material free for the taking.’ This, therefore, means that African data is collected without explicit permission and used to build profitable businesses, while Africans rarely benefit from the process or the output.

Beyond language models, AI is set to become the future of almost everything, from commerce to transportation to more immediate issues like health. AI-driven innovations are set to diagnose diseases non-invasively, predict epidemics and treat patients with personalized processes. Most of this, however, functions on colonial power structures.

The African Cancer Genome Registry (AGCR) project, led by Yemaachi Biotech, with the assistance of Roche and other international bodies, aims to map cancer genomics on a pan-African scale. Due to the lack of fair governance structures, this risks exporting Africa’s genomic goldmine in the name of collaborative science. Data-sharing systems, such as those suggested by the Global Alliance for Genomics and Health, backed by American, Canadian and European institutions, conveniently centralize ownership and monetization outside Africa.

This segues into an additional, often under-acknowledged danger in deploying AI systems trained primarily from data drawn from non-African populations. AI models in healthcare, policing, social welfare and other applications learn from examples and attempt to identify patterns, enabling them to predict or recommend actions based on the data they are exposed to. One example is biases in dermatology models, where most models, having been trained with predominantly light skin data, consistently misdiagnose dermatological conditions in dark-skinned Africans.

Other risks of non-indigenous models include unfair social profiling and biased financial risk assessments. Using AI that does not understand or represent African realities sustains stereotypes to the detriment of the communities involved. Unless AI is trained on localized and demographically balanced data sets for Africa, it will continue to be an extension of foreign power structures, policing, diagnosing, and deciding for a population it was never designed to serve.

DECOLONIZING DATA STRUCTURES

Decolonising AI in Africa is not merely about who owns the data; it is also about whose data builds the systems, whose experiences shape the algorithms, and ultimately, in whose interests they are working. Data Science Nigeria (DSN) shows how to build AI systems for Africa through its EqualyzAI project. DSN has developed AfroSLM 1.0, one of the first language models in Africa for the Yoruba and Hausa languages. They have also created speech models that work well with Nigerian English accents by recruiting and adequately compensating Africans to provide the voice data needed. This is an example of what AI in Africa could look like without colonial influence.

For Africans to have data sovereignty, it must be declared a constitutional right in our respective nations. We need robust data protection laws and truly local data centres that serve African economic and geopolitical interests. The Nigerian 2023 Data Protection Act and Kenya’s 2019 counterpart, the Data Protection Act, are steps in the right direction. They must, however, be backed by true enforcement.

The funding model also needs to be changed. African AI startups and research collectives should not always find themselves dependent on foreign grants and venture capital that dictate the terms of innovation through Western-centric metrics.  Governments, diaspora investors and African development banks should seize opportunities to fund homegrown AI models.

Africa urgently needs a strong, continent-wide data governance system. The continent lacks an enforceable standard data-governance system, meaning that each country negotiates with foreign giants separately. A Digital Sovereignty Protocol under the African Union could establish standard rules on data ownership, data retention and cross-border use, thereby providing Africa with a stronger voice and safeguarding citizens’ digital rights.

Finally, the AI education pipeline needs to be decolonized. African universities and training programs must move beyond being mere consumers of Western AI models to creators of context-specific, Afrocentric AI solutions. Just as the foundational colonization of Africa’s lands, labour and resources shaped the geopolitical maps we inherited today, so will the ongoing scramble for African data shape the business future of the continent.

This isn’t a philosophical debate or an academic exercise. It’s a political mission, an ethical imperative and an existential battle all at once. The data we have in hand tells us one unquestionable fact: those who control the data control the narrative—and those who control the narrative control the future⎈

*Nafisat Abdulsalam emerged in third place in the 2025 edition of The Republic Student Writing Competition, in partnership with the Open Society Foundation.