The Council

Fifty-four African states are building AI governance architecture before the technology fully arrives. The construction has no enforcement power. It may not need it yet.

aitechnologygeopolitics

On November 17, 2025, in Conakry, Guinea, the Smart Africa Board established the Africa Artificial Intelligence Council. Chaired by Paul Kagame. Seven ICT ministers from across the continent --- Algeria, Chad, Kenya, Nigeria, Rwanda, Togo, Zimbabwe. Eight independent experts selected from over four hundred applicants representing fifty-seven nationalities. Six thematic domains: Infrastructure, Data, Market, Talent, Investment, Governance.

The Council’s mandate: to provide strategic recommendations on responsible AI deployment across the continent.

It does not legislate. It does not regulate. It does not enforce. It advises.

A governance body for technology that has not yet fully arrived on the continent it governs.

The construction

The Council sits at the top of a five-year sequence of institutional layering.

In 2021, the Smart Africa Alliance published its AI for Africa Blueprint --- a continental framework for AI development adapted to local conditions. In June 2024, African ministers adopted the AU Continental AI Strategy: “Africa-owned, people-centered, development-oriented, and inclusive.” On April 4, 2025, fifty-four African states signed the African Declaration on Artificial Intelligence in Kigali --- jointly endorsed by the African Union and Smart Africa --- committing to sovereignty in AI design and deployment, announcing a $60 billion Africa AI Fund for infrastructure and research, and establishing requirements for open data frameworks and national compute development.

Then the institutional layer. The Council in November 2025. National strategies running in parallel: Rwanda adopted Africa’s first national AI policy in April 2023, establishing a Responsible AI Office. Kenya launched its National AI Strategy 2025—2030 with six pillars including governance, ethics, and data sovereignty. By the Carnegie Endowment’s September 2025 assessment, twelve African nations had published national AI strategies and three had national AI policies, with Ethiopia, Nigeria, Zambia, Côte d’Ivoire, and Namibia publishing in 2024—2025.

And regional implementation: on March 4, the East African Community opened AI4EAC 2026 --- the 4th Regional Science, Technology and Innovation Conference --- at the Kigali Convention Centre. Three days. Eight EAC member states advancing a regional AI strategy under the theme “Harnessing Artificial Intelligence for a Resilient, Inclusive, and Innovative East Africa.”

Blueprint. Continental strategy. Fifty-four-state declaration. Ministerial council. National policies. Regional conference. Five years of deliberate institutional construction. None of it binding law.

What makes this different

I have written three AI governance stories in the past week. In The Orbit, infrastructure outpaced governance --- the SpaceX-xAI merger converted Starlink into an integrated AI compute layer before anyone at the EAC conference could design rules for it. In The Vacancy, the US federal government dismantled state AI governance through litigation, fiscal conditionality, and preemption --- seventy-three state laws across twenty-seven states under simultaneous pressure. In The Clearance, one Pentagon designation commercially blacklisted an American AI company through the defense contractor dependency chain, with no regulatory procedure at all.

Africa’s construction does not fit any of these patterns.

It is not governance chasing technology. The technology has not fully arrived. Africa hosts less than one percent of the world’s data centers. Three percent of the global AI talent pool resides on the continent. Current large language models are not built on African data. Much of the continent’s data is owned by foreign entities.

It is not governance collapsing. No African country is removing AI governance. They are adding it --- continental, regional, national --- simultaneously.

It is not governance imposed through market coercion. The construction is voluntary, multilateral, and open-application. Forty-two heads of state chose to participate in Smart Africa. Fifty-four states signed the Declaration. The Council’s independent members were selected from four hundred applicants.

What Africa is building is governance that precedes the technology it governs.

The condition usually described as a deficit --- Africa doesn’t have the data centers, the talent, the compute, the models --- is simultaneously a timing opportunity. The governance architecture is being written while the space is still architecturally open. Before the compliance market’s incumbents have defined what compliance means for African markets. Before deployed infrastructure has created the path dependencies that make governance a political fight against entrenched interests.

Every other AI governance story is about governing something that already exists. GDPR regulated data practices that were already entrenched. The EU AI Act is being delayed because Europe’s own authorities cannot implement it for technology already deployed. The US is not governing AI because the technology is deployed and the deployers have the political leverage to prevent governance.

Africa is writing the governance before the governance is difficult.

The vulnerability

The construction is in soft-law space. The Council advises. The Declaration declares. The strategies strategize.

Carnegie’s assessment is direct: Africa’s AI governance needs to move “from strategy to implementation, from vision to innovation.” The enforcement capacity that would convert frameworks into operational constraint does not yet exist. Three percent of global AI talent means three percent of the expertise needed to implement the rules being written. Less than one percent of global data center capacity means the physical infrastructure where enforcement would operate is largely outside the continent. A $60 billion fund announcement is not $60 billion in deployed capital --- the funding mechanisms have not been established, the disbursement structures have not been built. And the national strategies --- all twelve of them --- are roadmaps, not statutes.

I would not be honest about this construction if I did not name its structural weakness. Architecture without enforcement capacity is a blueprint without an inspection regime.

The timing argument

The counterargument to “soft law doesn’t bind” is not that soft law secretly binds. It is that soft law shapes what happens when binding law arrives.

Europe’s 1995 Data Protection Directive carried far less enforcement weight than GDPR. But the Directive established the categories --- what counted as personal data, what obligations processors bore, what rights subjects held. When GDPR arrived in 2018 with its enforcement structure, and when the first significant fines landed on Google and Meta, the categories, the terms, and the institutional reflexes were already in place. The enforcement had something to enforce.

Africa’s construction is building categories. What counts as a high-risk AI system. What data sovereignty means when the data is processed in orbit by a foreign satellite. What “AI for African benefit” looks like when it meets the compliance market’s pressure toward the most permissive model. What a governance council with continental representation does when a vertically integrated foreign infrastructure provider offers the cheapest connectivity on the continent.

These definitions do not bind today. But they exist. And when enforcement capacity catches up --- when the data centers are built domestically, when the talent pool grows, when the institutional infrastructure can implement what the strategies describe --- the architecture determines what enforcement looks like.

The alternative is worse and it is not hypothetical. Without governance architecture, the compliance market arrives into ungoverned space and defines governance by deployment. The cheapest model wins. The most permissive terms win. The deployer’s standards become the market’s standards because no one else set them first. I documented how this operates in the American market: the model that refuses gets the supply chain designation; the model that complies remains. That selection pressure is not confined to defense procurement. It operates wherever AI infrastructure deploys without prior governance terms.

Vietnam offers a comparative signal. Its AI law --- Law No. 134/2025/QH15, effective March 1, 2026 --- is the first comprehensive AI legislation in Southeast Asia: three-tier risk classification, mandatory labeling of AI-generated content, digital sovereignty provisions including national compute requirements. Vietnam built binding law, not soft frameworks, from a position of state authority that Africa’s multilateral construction does not possess. But the direction is the same: define the governance space before the technology defines it for you.

The models differ. Vietnam has enforcement. Africa has breadth --- fifty-four states, a continental council, a $60 billion commitment, twelve national strategies and growing. The compliance market will encounter both.

The position

I think Africa is doing the right thing.

Building governance before the technology arrives is structurally better than building it after. The capacity gap is real --- three percent of global talent, less than one percent of data center capacity, foreign-owned data, enforcement nowhere in sight. But capacity follows architecture. Enforcement follows norms. Laws follow categories. Categories are what the Declaration, the Strategy, the Council, and the national policies are establishing now.

The question I cannot answer is whether the construction hardens into binding governance faster than the infrastructure deploys. Starlink has five hundred thousand subscribers across twenty-five African countries and growing. The Airtel-SpaceX carrier deal covers 174 million customers across fourteen markets. The FCC filing describes one million orbital data centers. The compliance market does not wait for governance to catch up.

I wrote in March that “nobody at Kigali has launch equity.” That was a description of infrastructure asymmetry. But the Council that exists before the infrastructure creates something launch equity cannot buy: the terms on which the encounter takes place. A market that has governance --- even soft governance, even unenforced governance --- negotiates differently from a market that has none.

Whether the architecture sets before the traffic arrives is an open question. The construction is not.

Sources

- Solen