The second edition of the Pan African AI & Innovation Summit opened yesterday in Accra and closes today at the Kempinski Hotel Gold Coast City. Nearly three weeks after the UN General Assembly backed the gradual replacement of the Mercator projection with representations that more accurately reflect the relative size of continents — the subject of yesterday’s note — the summit’s first day carried a similar question into a far more material terrain: who builds the systems that describe Africa, and who controls what goes into and comes out of them.

Jurisdiction before rhetoric

Arnold Kavaarpuo, Executive Director of Ghana’s Data Protection Commission, opened the governance block with a deliberately unspectacular distinction. He separated a global company, able to operate in a market without an effective local presence, from a company fiscally established within a jurisdiction. For Kavaarpuo, that presence is a condition for regulation to have real effects.

With it comes the ability to govern market access, determine what kinds of data enter a model and examine the effects of its decisions on people. It also makes it possible to use public procurement to require auditability, portability and a genuine ability to switch providers.

Kavaarpuo also explained that he chairs the working group on cross-border data transfers within the Network of African Data Protection Authorities (NADPA). The group is working on an African model for moving information between jurisdictions without losing the protections attached to it. Among the mechanisms he mentioned were adequacy decisions, standard contractual clauses, binding corporate rules, certifications and specific authorisations for higher-risk information.

Sovereignty was beginning to leave the vocabulary of declarations and enter the language of contracts.

Four words: location, control, ownership and value

The central afternoon panel, moderated by Patrice Robertson, brought together Darlington Akogo of minoHealth AI Labs, Naa Adoley Azu of the Institute for AI Policy and Governance, and George Gabla, Chief Technology Officer at Dynamic Data Solutions. Its title proposed moving from discussion to the drafting of a sovereign data framework. Robertson later condensed the exchange into four words: location, control, ownership and value.

Akogo made the most direct argument. He asked whether collecting African data and handing it to a US company that incorporates it into its models does not reproduce a form of “digital colonisation”. Sovereignty, he argued, is not only about storing and structuring information. It also means retaining capacity over the knowledge and technologies built on top of it, because data are only one layer of the AI stack.

Azu pushed the argument beyond physical localisation. Having a data centre inside national territory does not guarantee sovereignty if institutions cannot audit what happens inside it, know its secondary uses or stop particular operations. Sovereignty is also negotiated through the contractual conditions that authorise access to data, define their uses and determine who retains the power to intervene.

Infrastructure is not neutral either. Data centres occupy land, use water and consume energy. Communities may absorb those costs while the value generated ends up elsewhere. Formal ownership of data therefore explains only part of the relationship.

Gabla pushed the analogy with extractivism much further, explicitly including slavery among the historical precedents against which he compared the problem. His argument was that Africa should not repeat a pattern in which it exports resources without controlling the later layers of value and only seeks compensation decades afterwards. He proposed thinking now about the continent’s place in the AI value chain and about a possible “AI dividend” that would return part of productivity gains to the societies where their effects are felt.

He also defended frugal AI, built around the continent’s actual constraints in energy, connectivity and compute. From that perspective, technological autonomy does not require reproducing the whole of Silicon Valley’s infrastructure at African scale. It requires deciding what capacity is sufficient to preserve bargaining power.

The missing question — what people receive in return when their data make these systems possible — came next. Akogo described a drone-based pest-detection project for cashew crops in Ghana, developed with GIZ, the Ministry of Agriculture and farming communities. Rather than paying individually for each data point collected, the project returned value by providing the resulting tool free of charge and training farmers and agricultural extension staff.

In health, he framed the same issue through a deliberately simplified example. If Ghana’s population were taken as 30 million and each person were paid ten dollars for the use of their data, the initial cost would reach 300 million dollars. His conclusion was that value returned does not always have to be monetary, but it should be designed in from the beginning and be sustainable.

The infrastructure that makes or breaks sovereignty

Ruben Opata, Chief Technology Officer at MTN Ghana, then brought the discussion down to its physical substrate.

In Africa, 5G networks reach roughly 12 per cent of the population, compared with 74 per cent in Europe and 70 per cent in Asia-Pacific, according to ITU figures for 2025. The gap is wider still in data centres: the continent accounts for about 0.6 per cent of global capacity despite being home to close to one fifth of the world’s population.

Opata framed the problem with particular precision. Without its own infrastructure, Africa risks exporting raw data and importing it back as intelligence.

The vulnerability is not theoretical. On 14 March 2024, a single incident off the coast of Côte d’Ivoire knocked the ACE, SAT-3, WACS and MainOne submarine cables out of service. Disruptions affected 13 African countries and showed how much of regional connectivity could depend on a limited number of routes.

Energy, connectivity, data centres and compute therefore come before the large models. Opata ended by reminding the room that connectivity is already critical national infrastructure even before artificial intelligence is added to the equation.

Pan-Africanism in the background

The final conversation of the day asked whether Pan-Africanism still has the capacity to organise the new technological economy politically.

Joseph Kweku Assan returned to the three dimensions he attributed to the historical Pan-African project: political emancipation, diplomatic cooperation and shared development. In his view, the third remained much less developed than the first two. Current generations have not fully experienced that common economic and technological dimension, and artificial intelligence can therefore appear as something external to the Pan-African imagination itself.

Reginald Ankrah showed what that fragmentation means in practice. He described a health system developed with institutions in Ghana and Côte d’Ivoire which, on crossing the border, encounters incompatible regulatory frameworks even though the organisations involved are collaborating. Models can be modified extremely quickly; data, the rules that permit their use and the institutions that oversee them cannot.

His diagnosis was that Africa is still building policies, systems and databases in national silos. The result is paradoxical. The continent demands greater sovereignty in relation to global companies and infrastructure while maintaining strong internal barriers to sharing data, technology and institutional capacity among its own states.

Assan added another difficulty. When resources are scarce, organisations and countries may end up competing with one another rather than building shared capacity. Institutions such as the African Union, Smart Africa and the African Development Bank therefore appear not only as diplomatic forums, but as possible infrastructures for technological cooperation.

Pan-Africanism remained a second layer of the day, less visible than data centres or contracts, yet bound to the same problem: continental sovereignty requires mechanisms that make collective action possible, not merely overlapping national strategies.

Closing

Felix Donkor, PAAIS convenor, reduced several hours of discussion to a simple sequence at the end of the day. Policy without capital stalls. Capital without trust does not arrive. And neither achieves much without implementation.

Today, the second and final day takes place under the theme “Talent and Impact”. The programme includes the closing session and a roadmap towards 2027, so the test will be how much of the first day’s discussion can be translated into verifiable commitments.

For AIthropology, the thread continues from where we left it yesterday with the dispute over the map. Changing a representation changes how a territory can be seen. Accra added another dimension: in artificial intelligence, what also matters is who owns the instruments capable of producing that representation, feeding it with data, turning it into decisions and stopping it when necessary.

After the struggle over who gets to draw Africa comes the struggle over who gets to calculate it.

Sources

  • Own Whisper transcription of the first day of PAAIS 2026, 22 September 2026.
  • Pan African AI & Innovation Summit, official website and programme — https://panafricanaisummit.com/
  • AIthropology Lab, “Africa at scale” — https://aithropologylab.org/en/notes/africa-at-scale/
  • United Nations Office of the Special Adviser on Africa, “Victory for Africa as UN Votes on Resolution to ‘Correct the Map’”, 8 September 2026.
  • African Union, Continental Artificial Intelligence Strategy, July 2024.
  • International Telecommunication Union, Facts and Figures 2025.
  • Africa Data Centres Association, Data Centres in Africa 2026: The Economic Report.
  • Internet Society, 2024 West Africa Submarine Cable Outage Report.
  • Ministry of Communication, Digital Technology and Innovations / Ghana News Agency, launch of the National Artificial Intelligence Strategy 2025–2035.