There is a pattern I have noticed across every environment where I have worked to deploy technology at scale, whether building data pipelines for research institutions, designing operational systems for cross-border teams, or co-founding an EdTech platform serving hundreds of specialists across West Africa. The technical architecture rarely fails first. What fails first is the governance layer. The rules, the accountability structures, the shared understanding of what the system is for, whom it serves, and what happens when it goes wrong.
This observation has become more urgent, not less, as artificial intelligence moves from experimental to infrastructural across the African continent. And it raises a question that the region’s technology leaders, founders, and policymakers have not yet answered with sufficient seriousness: who is writing the rules that will govern AI in Africa, and are those rules written for us, or about us?
The Governance gap nobody is talking about loudly enough
Africa is not absent from the global AI conversation. Nigerian fintech platforms are deploying machine learning for credit scoring. Kenyan agrictech startups are using predictive models to optimise crop yields. South African healthcare systems are piloting AI diagnostics. The continent is building and building fast.
But governance has not kept pace with deployment. Most African organisations scaling AI products today are operating under one of two frameworks, and both are problematic.
The first is borrowed governance: adopting the EU AI Act, ISO standards, or US federal guidelines wholesale, without meaningful adaptation to local legal systems, data realities, or social contexts. These frameworks were designed in environments with robust data infrastructure, high digital literacy, high institutional trust, and legal systems capable of enforcing complex compliance obligations. Importing them into contexts where those conditions do not yet fully exist does not produce safety. It creates the appearance of safety-checkbox compliance that satisfies international partners while leaving local communities entirely unprotected.
The second is governance absence: deploying AI systems with no formal framework at all, relying instead on the good intentions of founders and the informal norms of small teams. This is understandable in early-stage startups where speed is survival. It becomes dangerous at scale.
At Vision AI Consortium, where I lead global operations, one of my core responsibilities is ensuring that as we deploy AI architecture across emerging and established markets, our governance does not become an afterthought retrofitted onto a system already in motion. I have seen, firsthand, how difficult that retrofit becomes once a system has users, dependencies, and commercial momentum behind it. The time to build governance is before scale, not after.
Why imported frameworks fall short
The limitations of borrowed governance are not abstract. They manifest in concrete, consequential ways.
Consider data. European AI governance frameworks are built on the assumption of GDPR-compliant data ecosystems, clearly defined data subjects, documented consent, enforceable rights of access and erasure. Across much of Africa, data infrastructure does not yet operate this way. Informal data collection, low digital literacy among end users, and fragmented regulatory environments mean that GDPR-derived principles, applied without adaptation, either cannot be meaningfully implemented or create compliance burdens that disadvantage local innovators competing against better-resourced international firms.
Consider cultural context. AI systems trained predominantly on Western datasets carry embedded assumptions about language, behaviour, identity, and risk that do not translate cleanly across African contexts. A credit scoring model validated in the United Kingdom will encode different assumptions about financial behaviour than one designed for the realities of Lagos or Accra. Governance frameworks that do not account for this, that do not mandate local validation, community consultation, and contextual auditing, will not catch the harms that matter most to African users.
Consider power dynamics. When African organisations adopt governance frameworks written by others, they also adopt the priorities of others. The EU AI Act is, among other things, a tool of European industrial policy designed in part to shape global AI norms in ways that favour European interests. There is nothing wrong with that; every major governance framework reflects the priorities of those who write it. The problem is when the frameworks of one region become the de facto global standard, not because they are universally appropriate, but because no alternative exists with sufficient institutional weight behind it.
What KnowBaze taught me about structural design
When I co-founded KnowBaze, the goal was straightforward in principle and complex in practice: build a structured pathway for people transitioning into technology, particularly women and underrepresented communities, in a context where existing resources were designed for different learners in different environments.
The lesson I took from that experience applies directly to AI governance. You cannot simply take a curriculum designed for learners in San Francisco or London, translate it into Pidgin, and call it localisation. The underlying assumptions about prior knowledge, about available tools, about what a career in tech looks like and who it is accessible to all require fundamental rethinking, not surface adaptation.
The same is true for governance frameworks. Localisation is not translation. It is a redesign, grounded in the specific realities, risks, and values of the communities the framework is meant to protect.
Building KnowBaze to serve 600 specialists and counting taught me that structural design done right is slower, harder, and more collaborative than importing an existing model. It also taught me that it is the only approach that works at scale because the communities you are designing for can tell, immediately, whether something was built for them or simply adjusted for them. The difference matters, and it shapes whether people trust the system enough to engage with it fully.
The case for African-led AI governance
None of this is an argument for isolationism or for rejecting international standards wholesale. Global interoperability matters. African AI systems will need to operate across borders, integrate with international platforms, and satisfy the governance expectations of global partners and investors. That reality is not going away.
The argument is more precise: Africa needs to be a co-author of global AI governance norms, not merely a recipient of them. And to be a co-author with credibility and leverage, the continent needs to develop its own frameworks built from African realities, informed by African expertise, and validated through African institutional processes that can then enter genuine dialogue with global standards.
This is already beginning to happen. The African Union’s Continental AI Strategy, adopted in 2024, represents a meaningful step toward continental coordination. Several national governments, including Rwanda, Kenya, and Egypt, have developed or are developing AI-specific policy frameworks. Civil society organisations and research institutions across the continent are contributing to the technical and ethical dimensions of this conversation.
But the private sector, the founders, the CTOs, the operators building and scaling AI products, have been largely absent from this governance conversation. That absence is both understandable and costly. The organisations with the most direct knowledge of how AI systems behave in African contexts, what risks they generate, and what governance mechanisms are practically implementable are the same organisations currently treating governance as a compliance obligation rather than a design principle.
That needs to change.
A practical starting point
I am not suggesting that every African startup needs a dedicated AI ethics board or a hundred-page governance policy. The practical starting point is more modest and more achievable.
It begins with acknowledging that the question “Is this system working?” is incomplete without asking “working for whom, and at whose expense?” It continues with building feedback mechanisms, real ones, not checkbox surveys that surface the experiences of the communities most affected by AI deployment. It extends to investing in local auditing capacity: the technical expertise to evaluate AI systems against contextually relevant standards, rather than relying exclusively on international certification bodies with no presence in the communities they are certifying.
And it requires the organisations with the most to gain from getting this right, the platforms scaling across multiple African markets, the consortium structures building AI infrastructure for the continent, to take governance seriously enough to resource it properly, not as a legal obligation but as a foundational design choice.
At Vision AI Consortium, this is the work we are doing. Not perfectly, and not completely, the field is too young and too complex for anyone to claim that, but deliberately, with the conviction that the governance architecture we build now will shape what AI looks like in Africa for the next two decades.
The window is open, but not indefinitely
Global AI governance norms are being written now. The frameworks being developed, debated, and institutionalised in the next three to five years will calcify into standards that are extremely difficult to revise once they achieve international consensus. Africa has a window, not a wide one, but a real one to shape those standards rather than inherit them.
That requires African technologists, policymakers, and operators to show up in the governance conversation with the same energy and sophistication they bring to product development and market expansion. It requires treating AI governance not as a constraint on innovation but as the infrastructure that makes innovation trustworthy enough to scale.
Because ultimately, the question is not whether Africa will be governed by AI frameworks. It will be.The only question is whether Africans will have written them.
Chioma Nneka Enyinnah is Chief Operations Officer at Vision AI Consortium and Co-founder of KnowBaze, an EdTech platform building structured technology pathways for underrepresented communities across West Africa. She specialises in AI governance, digital transformation, and operational systems design across emerging and global markets.
