The African Union has a habit of producing documents that sound like answers and function like questions. Its Continental AI Strategy is, so far, the latest iteration of this tradition — a serious, well-intentioned framework that describes a future Africa has not yet built the institutional capacity to reach.
That is not a criticism of the strategy’s authors. It is an observation about the structural conditions African governance operates within, and about what happens when the ambition of a document outpaces the resources, authority, and political will available to implement it.
What the AU’s AI Strategy Actually Says
The AU’s AI position is not contained in a single document but distributed across three interconnected frameworks: the Data Policy Framework adopted in 2022, the Digital Transformation Strategy for Africa, and the Smart Africa Model AI Policy Framework developed through the AU’s technology agency. Together, they constitute a coherent vision. The continent should develop AI capabilities locally, not import solutions from external providers. Member states should adopt common data governance principles to prevent the regulatory arbitrage that undermines continental trust. African voices should participate in setting global AI norms, not simply receive them from Geneva, Brussels, or Washington.
These are the right positions. The problem is the distance between a position and a policy, and the even larger distance between a policy and an enforcement regime with teeth.
Smart Africa — the AU’s primary technology delivery arm — has been active. Its Model AI Policy Framework, launched in Geneva in 2025, gives the 70% of African nations without national AI strategies a template to work from. The framework acknowledges the continent’s infrastructure reality explicitly: that African AI developers are forced to train models on foreign compute, store sensitive data abroad, and operate inside regulatory regimes built for foreign jurisdictions. This is meaningful. But a framework that names a problem is not a mechanism that solves it.
The Case for the Status Quo
There is a reasonable defence of the AU’s approach. Critics who demand binding continental AI legislation are ignoring the political economy of multilateral governance. The AU has 55 member states with wildly different levels of digital infrastructure, state capacity, and political systems. Rwanda has a functional AI Policy with tiered risk classification. South Sudan has more pressing problems than AI governance. Any binding instrument accommodating both will inevitably be written to the level of the weaker governance environment — rendering it useless for advanced economies and irrelevant for fragile states.
The Model AI Policy Framework’s voluntary nature is, by this reading, a feature. It creates a floor without imposing a ceiling. Countries that can move faster — Rwanda, South Africa, Kenya — can build on it. Smart Africa’s Community of Practice model also has genuine merit. Peer learning across policy practitioners is how governance capacity gets built in low-resource environments. When Namibia’s Ministry of ICT wants to know how to structure an AI risk classification system, talking to Rwanda’s ICT Chamber costs less and transfers more contextually relevant knowledge than hiring a European consultancy.
The incrementalist case is this: reach for binding instruments too early and you get instruments without implementation.
Why Incrementalism Has a Deadline
The counter-argument is simpler. AI systems are being deployed on African populations right now. Credit-scoring algorithms are deciding who gets loans in Lagos, Nairobi, and Accra. Facial recognition tools operate at borders. Agricultural AI is influencing crop decisions for smallholder farmers across the Sahel. Recruitment AI is filtering job applications at scale.
Every month that a continental governance framework exists only as aspiration is a month during which consequential AI systems operate without meaningful accountability. The incrementalist argument assumes governance can catch up to deployment. That assumption is being tested in real time, and the early results are not encouraging.
The EU took seven years from its first AI regulation proposal to the AI Act’s enforcement phase beginning in 2024. Africa started its continental AI conversation later and has fewer institutional resources to accelerate. If the trajectory holds, binding continental AI governance mechanisms are a decade away at minimum. By then, the systems that need governing will be a generation old and protected by the commercial interests that built them. This is the fragility inside the AU’s continental AI strategy. It is not strategically wrong. It is temporally inadequate.
What Would Make It Work
Three changes would materially improve the continental AI governance picture. None of them require a new treaty.
The AU should designate sectoral AI governance leads — specific bodies with named responsibility for AI in financial services, healthcare, and agriculture. These are the three sectors where AI deployment in Africa is deepest and accountability gaps are most consequential. Naming the regulators responsible for AI in those sectors, even on an advisory basis, creates accountability chains that currently do not exist.
The Smart Africa Community of Practice should be resourced to produce binding guidance documents, not just peer-learning sessions. The difference between a guidance document and a position paper is enforceability. If the AU cannot bind member states directly, it can create reference standards that national regulators adopt by choice — the way the Basel Accords work in banking, or the FATF standards in anti-money laundering. Soft law is better than no law.
The AU also needs to connect AI governance to its existing trade infrastructure. The African Continental Free Trade Area is the most significant economic policy project on the continent. If AfCFTA’s digital trade protocols include AI governance requirements — even light-touch ones — then every company seeking to operate across African markets faces a continental compliance floor. Trade frameworks create incentives that voluntary policy frameworks cannot.
The Credibility Question
There is a reputational dimension to this that goes beyond policy mechanics. Africa’s technology ecosystem has spent a decade arguing, with increasing credibility, that the continent is a serious market, a serious innovation environment, and a serious voice in global tech governance. African tech funding crossed $3 billion in 2025, with AI tools embedded across the most consequential sectors in the economy. The AI governance moment is a test of whether that seriousness extends to self-regulation.
The EU got its AI Act not because it was easy but because the political cost of inaction — regulatory fragmentation, citizen distrust, exclusion from global norm-setting — was judged higher than the cost of acting. Africa’s continental bodies have not yet reached that calculation. They should.
The AU’s AI strategy is bold on paper. Making it functional in practice requires the same urgency the continent brings to its economic ambitions — and an honest reckoning with the distance between the two.
