Kenya’s AI Divide: How Colonial Extractivism and Silicon Savannah Hype Mask Structural Exclusion in Innovation

Kenya’s innovation ecosystem could draw from indigenous knowledge systems like harambee (communal labour) or age-set systems (intergenerational knowledge transfer) to design AI that prioritises collective welfare over profit. Traditional ecological knowledge, such as the Maasai’s seasonal migration patterns or the Pokot’s drought-resistant farming techniques, offers data-driven insights for climate-adaptive AI—yet these are rarely integrated into tech solutions. The erasure of such systems reflects a broader devaluation of African epistemologies in favour of Silicon Valley’s 'move fast and break things' ethos.

Kenya’s tech sector is built on colonial infrastructures of extraction, from the railway lines that facilitated cash-crop exports to the post-independence IMF-mandated structural adjustment programmes that privatised public services. The Silicon Savannah narrative mirrors earlier 'African miracle' stories (e.g., mobile money in the 2000s) that promised liberation but delivered corporate control—M-Pesa, for instance, was co-opted by Safaricom to deepen financial surveillance. Historical parallels abound in other extractive economies (e.g., South Africa’s platinum mines, Nigeria’s oil fields), where 'innovation' masked resource plunder.

While Silicon Valley frames 'disruption' as inherently progressive, many non-Western cultures view technological change through the lens of relational ethics—prioritising harmony over efficiency. In India, the concept of 'jugaad' (frugal innovation) emphasises resourcefulness within constraints, offering a counterpoint to Kenya’s venture-capital-driven model. Latin American 'buen vivir' philosophies similarly reject growth-at-all-costs paradigms, advocating for technologies that serve ecological and social well-being. These frameworks challenge the assumption that AI must be scaled globally to 'work for everyone.'

AI systems in Kenya are often trained on biased datasets (e.g., Nairobi’s informal settlements, rural agricultural data) that reflect historical inequities, leading to discriminatory outcomes in credit scoring, policing, or healthcare. Studies show that facial recognition systems perform poorly on darker-skinned populations, yet these tools are deployed in Kenyan airports and banks without validation. The 'move fast' approach ignores peer-reviewed evidence on algorithmic bias (e.g., Buolamwini & Gebru’s 2018 Gender Shades study) and the risks of reinforcing colonial gaze in computer vision systems.

Artists like Njideka Akunyili Crosby and musicians like Juliani critique Kenya’s tech hype through works that juxtapose Silicon Savannah promises with lived realities of inequality. Spiritual traditions, such as the Digo people’s reverence for the ocean as a sacred entity, clash with AI-driven marine conservation projects that treat ecosystems as extractable data. The Kikuyu concept of 'gĩkũyũ kĩa mĩthamo' (the wisdom of the elders) offers a framework for intergenerational knowledge transfer that could guide ethical AI development—yet this is dismissed as 'unscientific.'

Scenario modelling suggests Kenya’s AI sector could either become a hub for equitable, community-owned technologies or a dystopian testbed for surveillance capitalism, depending on governance choices. If current trends continue, AI will deepen precarity for gig workers (e.g., Uber drivers, content moderators) while enriching a tiny tech elite—mirroring patterns seen in China’s 'digital socialism' or India’s gig economy. Alternative futures include cooperative AI models (e.g., platform cooperativism) or decolonial AI designs that centre indigenous epistemologies and ecological limits.

Marginalised groups—informal workers, smallholder farmers, pastoralists, and disabled communities—are treated as 'data points' rather than stakeholders in AI design, their labour and knowledge commodified without consent. Women in Kenya’s informal settlements, who bear the brunt of digital exclusion, are rarely consulted in AI policy debates despite being primary users of mobile money and digital health tools. The framing of 'making AI work for everyone' assumes a homogenous 'everyone,' ignoring how class, gender, and disability shape access to and benefits from technology.