Africa’s AI Rulebook Is Being Written on Three Floors at Once
The African Union is drafting a model AI law, regional blocs are writing strategies and national parliaments are racing ahead with bills. The binding constraint on all three is not ambition but the state's capacity to draft, verify and enforce.
Africa’s approach to governing artificial intelligence is often described as a gap: few laws, thin regulators, models built elsewhere. September suggested otherwise. Within a few weeks the African Union consulted on continental rules, a regional bloc in the Horn of Africa began drafting its own framework, Kenya’s president took an AI coalition to the UN General Assembly and South Africa’s communications minister explained, in public, how his department’s flagship AI policy had collapsed under fabricated citations.
The activity is real. The problem is that it is happening on three floors at once, continental, regional and national, with no lift between them. Companies and governments that plan for a single African AI regime, or for none, will misread the next two years.
The top floor: a continent drafting by consensus
The African Union’s Continental AI Strategy was endorsed by its Executive Council in Accra in July 2024. Two years on, the work has moved from vision to plumbing. On 23 and 24 September the AU Commission’s Infrastructure and Energy Department held an open consultation on AI governance and regulation, according to TechAfrica News, with draft guidelines and a model law being prepared for ministerial review.
Souhila Amazouz, the commission’s principal coordinator for AI and digital policy, told participants that 17 African countries had now formally adopted national AI policies, up from eight when the strategy was being developed. That is progress, but it also means roughly two-thirds of the AU’s 55 members still have none. A model law is the AU’s usual answer to such unevenness: a template states can adopt, adapt or ignore.
Speakers pointed to the prize of acting together. Mehdi Snene, a senior adviser to the UN Secretary-General’s envoy on technology, noted that Africa’s 54 UN member states make up more than a quarter of the organisation’s membership. Angela Wamola of the GSMA, the mobile industry body, argued that sovereignty should mean strategic autonomy rather than each country pursuing technological isolation on its own.
The middle floor: blocs and coalitions
Regional blocs are writing their own documents. On 4 and 5 September the Intergovernmental Authority on Development (IGAD), whose members span the Horn of Africa and parts of East Africa, opened consultations in Addis Ababa on a Regional AI Framework and Strategy. The workshop reviewed an AI readiness assessment, validated country inputs and examined existing initiatives to avoid duplication. The bloc says the effort is aligned with both the AU strategy and the UN’s Global Digital Compact.
Kenya, meanwhile, is building a different kind of middle layer. At a roundtable on the margins of the UN General Assembly on 21 September, President William Ruto joined Finland’s president, Alexander Stubb, and Sierra Leone’s president, Julius Maada Bio, to discuss an “AI Middle Powers” initiative for countries that cannot build full AI stacks alone and want to pool computing power, data and talent. According to Innovation Village’s weekly digest, Kenya proposed agreeing common terms when buying AI models and compute, testing systems jointly and building datasets in local languages.
That last idea deserves more attention. Collective procurement would let smaller states negotiate on data residency, pricing and evaluation with suppliers that would otherwise deal with them one at a time.
The ground floor: parliaments in a hurry
National legislatures are not waiting for the floors above. Kenya’s Artificial Intelligence Bill, 2026, sponsored by the nominated senator Karen Nyamu, received its first reading on 2 April and was committed to the Senate’s ICT committee. It would create an Office of the AI Commissioner with powers to inspect premises, audit systems and levy fines, and sort AI into four risk tiers borrowed from the European Union’s AI Act. It would also criminalise deceptive deepfakes, with penalties of up to KES5m and two years in prison.
Critics are not opposed to rules so much as to sequencing. The law firm Bowmans has asked whether the bill was “proposed too soon”, noting that it leaves the detailed classification criteria to future regulations and may overlap with data-protection duties. Writing in Tech Policy Press, Liz Orembo argued that it bypasses Kenya’s National AI Strategy 2025–2030 and the consultations around it, and offers no exemption for satire, despite the use of AI tools in civic protest during the #RejectFinanceBill movement.
Nigeria shows the opposite risk: a law that is always about to arrive. In January Techpoint Africa reported that lawmakers aimed to pass the National Digital Economy and E-Governance Bill by the end of March, summarising its AI provisions as giving the regulator, NITDA, power to demand information and suspend unsafe systems, requiring annual impact assessments from developers and allowing fines of up to ₦10m or 2% of annual revenue. In July Senator Shuaib Salisu, who chairs the Senate’s ICT committee, said the bill had completed its Senate stages and was awaiting presidential assent. At the last public word, the country that set out to be among Africa’s first with comprehensive AI legislation was still waiting for a signature.
A model law can be copied in an afternoon; the capacity to check its footnotes takes years to build.
The real bottleneck is the state itself
South Africa offers the sharpest lesson. Its draft national AI policy, gazetted for comment in April, was withdrawn within weeks after it emerged that its reference list contained fictitious sources. Speaking at the GovTech conference in Durban on 21 September, the communications minister, Solly Malatsi, acknowledged that officials had not checked the citations. “We didn’t interrogate the bibliography,” he said, according to iAfrica.com. A redraft, led by an expert panel chaired by Professor Benjamin Rosman of the Wits MIND Institute, is now due before Cabinet by March 2027, Daily Maverick reported. In the meantime, iAfrica noted, the World Bank has left South Africa off its map of national AI strategies.
The episode is embarrassing, but its significance is structural. The continent’s most industrialised economy could not get a policy document through basic verification. Drafting AI rules that are workable, consistent with data-protection law and enforceable by a stretched regulator is harder still.
For decision-makers, three conclusions follow. First, the AU’s guidelines and model law will set the vocabulary of African AI regulation well before most national statutes take effect; firms and ministries that want a say should engage now, while the drafts are still soft. Second, the national picture will remain uneven and occasionally contradictory, as Kenya’s bill and strategy show, so compliance planning should be country by country, anchored in the data-protection regimes that are already enforced. Third, the most valuable investment governments can make is unglamorous: drafters and regulators able to scrutinise both the technology and their own paperwork.
As Girmaw Abebe Tadesse, a member of the UN’s new scientific panel on AI, put it to UN News in September, African countries need to share experience “instead of each starting from scratch”. Three floors of rule-making can work if someone builds the stairs.
Sources
- The AU’s Question: What 55 Countries Need to Agree on Before Africa Can Govern AI – TechAfrica News
- IGAD Launches Landmark Consultations on Regional AI Framework and Strategy – IGAD
- AI Africa Intelligence, 17–23 September 2026 – Innovation Village
- Africa needs its own AI solutions, not borrowed ones, says expert – UN News
- What Kenya’s AI Bill Gets Wrong About Political Expression – Tech Policy Press
- Kenya: Artificial Intelligence Bill, 2026 – Proposed too soon? – Bowmans
- Nigeria eyes passage of AI law to lead regulation push in Africa – Techpoint Africa
- Sen. Salisu highlights digital economy bill at T&A Legal AI summit – P.M. News
- Hallucination-free (hopefully) — new draft AI policy will be ready by March, says Malatsi – Daily Maverick
- ‘We Didn’t Interrogate the Bibliography’: Malatsi on the AI Policy Collapse – iAfrica.com
Discussion
No comments yet. Start the conversation.
