The African Union asked 55 countries this week how to govern artificial intelligence without splintering the continent. Any answer will depend on whether national regulators can enforce it. Africa AI governance is moving from strategy documents to draft rules, and the hard questions are now about delivery, not intent.
The AU Consultation
The AU’s Infrastructure and Energy Department opened a two-day open consultation on AI governance and regulation on September 23. It aims to put into practice the Continental AI Strategy, which the AU Executive Council endorsed in Accra in July 2024. According to TechAfrica News, the AU is preparing draft guidelines and a model law for ministerial review.
That sequence matters. A strategy sets direction, but a model law gives governments something concrete to adopt, amend or reject. The consultation is the point where broad principles start to meet the practical concerns of regulators, businesses and civil society.
Souhila Amazouz of the African Union Commission said countries with formal national AI policies rose from eight to 17 since the strategy launched, roughly one-third of African nations. Mehdi Snene of the UN Office for Digital and Emerging Technologies said 80% of African countries still lack a national strategy. The two figures do not reconcile, and accounts of the session offer no explanation.
The gap could reflect different definitions of what counts as a national AI policy or strategy, or different cut-off dates. Until the sources explain their methods, readers should treat both numbers with caution. What both figures agree on is the direction of the problem: most African countries are still without a settled national approach, which is exactly why a continental baseline is on the table.
A Baseline, Not a Single Rulebook
Angela Wamola, head of Africa at GSMA, argued for a shared floor rather than uniform statutes. “We’re not talking about imposing identical legislation,” she said. Countries would move toward interoperable rules at their own pace.
The distinction is important. Identical laws would be slow to negotiate and hard to fit onto very different legal systems and economies. A baseline lets countries with more mature frameworks keep moving, while giving others a template to work from. Interoperability, in this framing, means a company or public body could operate across borders without facing wholly incompatible requirements.
She also wants data centres, energy and connectivity treated as shared foundations, a case that echoes Smart Africa’s push to draft cross-border data exchange guidelines with 11 nations. The argument is that rules alone will not help if the infrastructure needed to run AI systems is patchy or unaffordable.
Wamola also said Kenya ranks first globally in active ChatGPT users, and that African data leaves the continent without producing local value. TechMoonshot could not independently verify the ranking. The second point, about data leaving the continent, is a policy argument rather than a statistic, and it explains why data governance sits near the centre of the discussion.
Snene pressed the external case. Africa’s 54 UN member states hold more than a quarter of the membership, he said, and should arrive with a few clear demands: equitable access to compute, data and talent, predictable financing, and a seat in standard-setting bodies. He urged delegations to agree a common position before the second UN Global Dialogue on AI Governance in New York.
The logic is one of bargaining strength. A bloc that speaks with one voice carries more weight than dozens of separate delegations, but only if its members settle on shared priorities first. That work has to happen before the New York meeting, not during it.
Implementation Is the Sticking Point
Practitioners worried most about delivery. Robert Ifeonu of the Central Bank of Nigeria feared the strategies would become well-written documents that institutions cannot use, and asked what a framework means for a central bank or tax authority. His question is a useful test for any AI rulebook: if a sector regulator cannot translate it into day-to-day practice, it will not change much.
Girmaw Abebe Tadesse of Microsoft’s AI for Good Lab noted that systems built in Lagos can fail in rural Nigeria or Rwanda if nobody tests them there. That is a reminder that AI governance is not only about legal text. It also covers testing, local data and the ability to catch failures before they reach the people who depend on the system.
According to TechAfrica News, the consultation named data governance, cross-border data flows, cybersecurity and sandbox testing as the areas where alignment matters most. Those are domains where data regulators already work, which makes their capacity central to Africa AI governance. Countries with established data protection authorities have a head start. Those without them will have to build that capacity while also absorbing new AI rules.
The AU’s model law will also land in a crowded field. Smart Africa launched a model AI policy framework in 2025 for the majority of countries that lacked one. Governments now have more than one template to choose from, and the AU will need to show how its model law fits with existing efforts rather than adding another layer. A baseline that regulators cannot enforce is a wish list.
What to Watch Next
Three developments will show whether the consultation turns into real progress:
- The AU’s draft guidelines reach ministers. The move from consultation to ministerial review is the first concrete step toward a model law.
- African delegations arrive in New York with one position. Agreement on access to compute, data, talent, financing and a voice in standard-setting bodies would strengthen the continent’s hand.
- National regulators show they can enforce what gets adopted. This is the test that decides whether any baseline changes outcomes on the ground.
