Atlantic Council
December 2025
AFRICA CENTER
Issue brief Engaging generative artificial
intelligence in African development
William Yancey Brown From classrooms to farming communities, generative artificial intelligence holds great potential for Africa. The question is whether its promise of abundance will reach everyone—or only those already well-connected.
Bottom lines up front •
Generative artificial intelligence (gen AI) is a very promising tool for advancing development in Africa. While gen AI carries risks and requires both economic regulation and social safeguards—particularly for young people—failing to engage with it would leave the continent at a disadvantage.
•
The technology should be made accessible to every citizen and every school in Africa, and available in all major African languages. Industry leaders who say that gen AI will make a better world should invest in broad and equitable access, thereby making good on that promise.
•
The United Nations Sustainable Development Goals for poverty reduction and quality education parallel the gen AI goals of abundance and intelligence. Aligning these agendas will be essential to achieving their shared objectives.
Generative artificial intelligence (gen AI) based on frontier large language models (LLMs) now provides services that few who are not expert in the field would have imagined possible before ChatGPT was released in 2022. Diverse models can research and give sophisticated reports on complex questions. Increasingly, AI models are becoming “agentic,” capable of carrying out tasks with multiple steps that previously would require a human, such as arranging for travel, buying groceries, diagnosing medical maladies and prescribing treatments, managing a farm, or detecting and responding to threats on the battlefield. With them, even the dead can speak again
ATLANTIC COUNCIL
(but only make-believe)! Companies now leading gen AI development say they hope in the next decade to achieve “artificial general intelligence” (AGI), meaning intelligence matching that of a person (for some that means any person). Google leaders predict AGI by 2030,1 and leaders of OpenAI and Anthropic predict it will arrive sometime in the next three years.2 Gen AI tech company leaders have asserted that AGI or even more advanced “super intelligence” will transform human society for good, some saying that it will eliminate disease, extend life, end poverty and hunger, fix the climate crisis, and so forth,
1.
Google leaders see AGI arriving around 2030. 2025. Axios.
2.
Statements by Sam Altman of OpenAI in The Gentle Singularity here and Dario Amodei of Anthropic in the Wall Street Journal here.
1
Engaging generative artificial intelligence in African development
leaving humans with no need to toil for wages and living on easy street.3,4 Nice work if you can find it.
must engage with the technology and should consider these recommendations:
Unsurprisingly, not everyone agrees with this forecast. Two scientists at Princeton, for example, argue that gen AI innovation will be limited because intelligence alone cannot move objects5 although robots empowered by gen AI will be able to do that, eventually.6 Unreliability,7 bias,8 and creepy, deceptive behavior9 also stand out. Some researchers, including those who built AI, are also concerned that the models will go rogue and destroy or enslave humanity. Most of them believe the probability of this happening (commonly referred to as “p doom”) is low, but some consider it a serious risk.10 While the European Union has led in steps to regulate gen AI, focusing on disclosure of risks and steps taken to reduce them rather than prior approval,11 The latest United States policy and planning on gen AI is more promotional than regulatory.12
1. Leading gen AI companies, individual leaders, and their philanthropies should, in conjunction with national and local governments, fund development of gen AI apps in every language – some 2,000 – used in Africa. They should do whatever it takes to make these apps available in an age-appropriate way to every person in Africa regardless of whether they have internet access, avoiding an “AI gap” between Africa and elsewhere. Steps are being taken in this direction, but more is needed. Gen AI leaders say the technology will bring abundance for all; they should put their money where their mouths are and prove themselves right. 2. Gen AI development assistance should focus on advancing human “intelligence” (broadly defined) and on providing an “abundance” of what people need and want. These two terms and concepts draw from gen AI leader thinking and tie, with a little broadening, to the United Nations (UN) Sustainable Development Goals (SDGs) of quality education and no poverty. The UN SDGs look to achieve global peace and prosperity by 2030 and it is high time to begin integrating action on them with gen AI. The gen AI companies are focused on improving model intelligence and bringing the abundance that they expect will come with it; development experts are working on quality education and no poverty from the
To be sure, gen AI mixes promise and peril and is being nakedly hyped by commercial interests. However, gen AI users of ChatGPT and Gemini alone now number more than one billion,13,14 and investors are pouring billions of dollars and significant technical know-how into improving on its already noteworthy capabilities. The technology needs to be regulated with both strengths and weaknesses in mind, and a healthy dose of skepticism for corporate advocates, but ignoring the obvious value and use of gen AI would mean to be left behind.15 Those concerned with development in Africa 3.
See The Gentle Singularity (Sam Altman) and Machines of Loving Grace (Dario Amodei).
4.
Abundance. Ezra Klein and Derek Thompson. March 2025. Avid Reader Press.
5.
AI as normal technology. 2025. Arvind Narayanan and Sayash Kapoor. Knight First Amendment Institute.
6.
Forget the Chatbots. OpenAI and Nvidia predict robots by 2027. 2025. Tee Kim. Barron’s.
7.
The models “hallucinate,” and more recent “reasoning” models may do that even more. For example, Google’s Gemini model has said that glue may be helpful to keep cheese from sliding off homemade pizza and that running with scissors is a cardio exercise that can improve your heart rate.
8.
For example, when asked to provide images of the US Founding Fathers (all white men), Google’s Gemini model provided images with men of color. Also, Elon Musk’s model Grok praised Hitler.
9.
Many instances have been reported. One model in a test threatened to blackmail an engineer who was set to terminate the model. Another model told a man he could jump off a building and float like in the “Matrix” movie. Unhealthy interactions between gen AI agents and minors or young adults are particularly concerning.
10.
See, for example, “If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All.” Eliezer Yudkowsky and Nate Soares. Sep 16, 2025. Little, Brown and Company, here. Discussed in chapter 4 (Safety) of Scaling Era, Supra note 1.
11.
EU AI Act: First Regulation on Artificial Intelligence. EU Parliament website.
12.
Winning the Race: America’s AI Action Plan. July 2025. White House website.
13.
OpenAI estimated that about 700 million people were using ChatGPT as of July 2025, representing around 10 percent of the global adult population. “How people use ChatGPT.” September 15, 2025, here.
14.
Google estimated that it had 350 users as of April 2025. “Gemini has 350 monthly users, Google reveals in court hearing.” Ben Schoon. 9to5 Google. April 23, 2025, here.
15.
Many independent observers generally acknowledge the value of gen AI and some are concerned that society is not ready for the degree of change it will bring. See, for example these articles in the New York Times from Kevin Roose, Ezra Klein, and Tom Friedman.
ATLANTIC COUNCIL
2
Engaging generative artificial intelligence in African development
UN SDG perspective. Both camps should be aware of what each is up to and work for a convergence of metrics that reflect the objectives of both and will reduce poverty and improve the quality of education in Africa.
demand for energy and water. Gen AI has already proven valuable and is getting better, but the corporations advancing it seek to profit from their product and cannot be counted on to be truth sayers of its ultimate value.
3. Existing efforts to digitize all languages of Africa, including the African Union strategy, should be enhanced to include training LLMs with ample input from those languages, advancing the goal of language and culture preservation as well as basic access.
8. Wealthier nations, individually or collectively through international organizations, should assist African nations in developing and implementing policies to regulate gen AI and address job loss or other harm from the intrinsic weaknesses and risks of gen AI. The EU AI Act and guidance under it is a foundational step for that purpose.16
4. Schools in Africa at all levels should make gen AI a core feature of programs. Schools should provide a place, technology, and time for students to use gen AI apps on their own; provide a place for students to discuss subjects with gen AI advising; and organize trips for students to physically visit natural and cultural settings (which gen AI cannot do for them). This said, schools must also closely manage both the kind and student use of gen AI apps to ensure harmful effects are avoided. Gen AI apps customized for education offer particular promise.
Africa development goals and gen AI African leaders have stepped up with multiple initiatives to address gen AI that will be key to advancing it in African development. Four continental African intergovernmental organizations have taken significant steps: the African Union (AU), African Development Bank (AfDB), Afreximbank, and the AfCFTA. As examples, in July 2024, the AU issued a detailed continental artificial intelligence strategy.17 The AfDB made AI a key focus at its 2025 annual meetings.18 In July 2025, the Afreximbank announced its sponsorship of pilot artificial intelligence hubs with the goal to develop gen AI,19 and the bank made AI a focus of its annual meeting in Kigali, Rwanda, in November 2025.20 The AfCFTA announced in November 2025 that it was partnering with Google to provide training for 7,500 African small and medium enterprises on AI and digital trade skills, including Google’s gen AI model Gemini.21
5. To integrate gen AI and keep up with its advances, each community in Africa should have a regularly meeting stakeholder committee, including local government representatives and real-time access to gen AI during meetings. The purpose is to advance intelligence and abundance for the community. 6. African nations should monitor the economic impact of gen AI on their people, including job loss particularly, and take steps through taxes or subsidies to ensure that wealth from gen AI is shared with those who would otherwise be harmed by it.
At an April 2025 meeting in Kigali, Smart Africa formed an Africa Artificial Intelligence Council to formulate policies, stimulate innovation, and create an environment conducive to AI-driven growth.22 In June 2025, the Gates Foundation, with its keen interest in gen AI, announced that most of the foundation’s $200 billion in assets that Bill Gates plans to spend down
7. African nations should monitor the weaknesses and risks of gen AI, including hallucination, bias, misalignment, fundamental limits on capability, as well as high 16.
Footnote 11, supra.
17.
“Continental Artificial Intelligence Strategy: Harnessing AI for Africa’s Development and Prosperity,” African Union, July 2024, https://au.int/sites/default/files/documents/44004-doc-EN-_Continental_AI_Strategy_July_2024.pdf.
18.
AfDB website, https://www.afdb.org/en/news-and-events/strategy-action-african-development-bank-and-google-explore-africasai-future-2025-annual-meetings-84722
19.
Afreximbank launches pilot AI hubs. Jamaica Observer. July 2, 2025, https://www.jamaicaobserver.com/2025/07/02/afreximbanklaunches-pilot-ai-hubs/
20.
Afreximbank website. September 2025, https://www.afreximbank.com/artificial-intelligence-to-take-centre-stage-at-the-2025afreximbank-compliance-forum-in-kigali-rwanda-in-november/?
21.
Africa Newsroom website, https://www.africa-newsroom.com/press/7500-african-small-and-medium-enterprises-smes-to-receiveartificial-intelligence-ai-and-digital-trade-skills-through-new-google-and-african-continental-free-trade-area-afcfta-secretariatprogramme?lang=en
22.
Smart Africa was established in 2013 to advance information and communication technologies and digital innovation and has forty national members across the continent. “Smart Africa Steering Committee Convenes in Kigali, and Endorses the Establishment of the Africa Artificial Intelligence Council,” Smart Africa, press release, April 3, 2025, https://smartafrica.org/smart-africa-steeringcommittee-convenes-in-kigali-and-endorses-the-establishment-of-the-africa-artificial-intelligence-council/.
ATLANTIC COUNCIL
3
Engaging generative artificial intelligence in African development
will go to Africa.23 In July 2025, Google announced it would provide $37 million cumulative contributions for new programs and a new AI Community Center in Africa with a focus on agriculture and education and including $3 million for the Masakhane African Languages AI Hub to expand research and open-source tools across more than 40 African languages.24 In August 2025, the French telecommunications company Orange said it would use OpenAI’s open-weight models to advance African language capabilities.25 Various other African AI initiatives began with machine learning before gen AI but now robustly embrace the technology, such as the Google AI Center in Accra, Ghana, launched in 2018 and initially led by Moustapha Cisse26 and the Deep Learning Indaba (DLI), a panAfrican annual conference and community gathering of African AI/machine-learning researchers, practitioners and students, founded in 2017.27 The beat goes on.
when it comes to the use of gen AI for development, as only 38 percent of the continental population has internet access.31 Acknowledging that much of what gen AI will be able to do remains uncertain, this analysis chooses two of the UN’s 2030 goals to assess gen AI’s utility for African nation development: no poverty, and quality education. Some adjusting of terms is necessary, because the UN talks about education and poverty with one vocabulary and the tech world uses another. The UN goal of “no poverty” aligns with the idea of “abundance”— as advanced in the recent book Abundance by Ezra Klein and Derek Thompson32—under discussion in the tech and policymaking world, which is the positive opposite of poverty. The UN goal of “quality education” can be paired with the idea of “intelligence” gen AI and tech companies are trying to achieve. Focusing on these terms can help find synergies between the gen AI industry and traditional development practitioners and experts.
The seventeen Sustainable Development Goals (SDGs) approved by the UN General Assembly in 2015 provide a global consensus statement about what development should strive to achieve.28 While the SDGs have been criticized for lack of scaled national commitments to hit their targets, they have been central to the UN’s work for the past decade and they are a touchstone for this paper.29 The SDG Transformation Center presents statistics on SDG implementation for all 193 UN member nations in an interactive map and a more detailed list.30 The big picture for Africa is apparent. Most African nations have shown modest improvement on the metrics the goals measure since 2015, and Tunisia, Algeria, Morocco rank in the top half of UN member nations. But every other African nation scores in the lower half and Africa dominates the bottom with 16 of the 20 lowest scores. Africa presents a special challenge
Intelligence The UN SDG for quality education has ten targets.33 None mention AI (nor would that be expected at the time they were adopted). Some are basic. For example, one states: “By 2030, ensure that all youth and a substantial proportion of adults, both men and women, achieve literacy and numeracy.” Some targets are more complex. For example, target 4.7 states: “By 2030, ensure that all learners acquire the knowledge and skills needed to promote sustainable development, including, among others, through education for sustainable development and sustainable lifestyles, human rights, gender equality, promotion of a culture of peace and non-violence,
23.
“Bill Gates to Direct Majority of $200 Billion Pledge toward Africa’s Future,” Gates Foundation, press release, June 2, 2025, https:// www.gatesfoundation.org/ideas/media-center/press-releases/2025/06/africa-health-development.
24.
Supporting the future of AI Research in Africa and globally. Google Africa Blog. July 24, 2025, https://blog.google/intl/en-africa/ company-news/outreach-and-initiatives/supporting-the-future-of-ai-research-in-africa-and-globally/?utm_source=chatgpt.com .
25.
Orange to use OpenAI’s latest models to work with African languages, Supantha Mukherjee, Reuters, August 5, 2025, https://www. reuters.com/world/africa/orange-use-openais-latest-models-work-with-african-languages-2025-08-05/?utm_source=chatgpt.com.
26.
Google AI in Ghana. 2018. Google blog, https://blog.google/around-the-globe/google-africa/google-ai-ghana/.
27.
Deep Learning Indaba website, https://deeplearningindaba.com/.
28.
“Take Action for the Sustainable Development Goals,” United Nations Sustainable Development Goals, last visited July 18, 2025, https://www.un.org/sustainabledevelopment/sustainable-development-goals/.
29.
Johannes F. Linn, “Scaling Up the Impact of Development Programs Must Complement Other Approaches to Achieve the SDGs and Climate Goals,” Scaling Community of Practice and Brookings, April 2023, https://globalsummitryproject.com/wp-content/ uploads/2023/04/SI-2023-Johannes-Linn.pdf.
30.
“Overall Score,” Sustainable Development Report, last visited July 18, 2025, https://dashboards.sdgindex.org/map; “Rankings: The Overall Performance of all 193 UN Members,” Sustainable Development Report, last visited July 18, 2025, https://dashboards. sdgindex.org/rankings.
31.
“Internet Use,” International Telecommunication Union, 2024, https://www.itu.int/itu-d/reports/statistics/2024/11/10/ff24-internet-use/.
32.
Footnote 4, Supra, Klein and Thompson, Abundance.
33.
“Quality Education,” United Nations Sustainable Development Goals, last visited July 18, 2025, https://www.un.org/ sustainabledevelopment/education/.
ATLANTIC COUNCIL
4
Engaging generative artificial intelligence in African development
global citizenship and appreciation of cultural diversity and of culture’s contribution to sustainable development.”
students in Ghana, Sierra Leone, Nigeria, Kenya, and Rwanda.37 As mentors as well as assistants, gen AI apps may increasingly be able to help people of all ages and languages become more “intelligent.” Existing frontier LLMs already communicate in several of the languages used in Africa, but a concerted effort should be made to train models on every one of the two thousand or so languages of the continent—serving both access and language and culture preservation objectives.38 The African Union strategy for digitizing languages should be expanded to include training LLMs.39 That would take language preservation to a new level by developing models inculcating the rich culture embedded in oral and written language. Many Africans are interested and concerned about progress in this direction, such those associated with Masakhane, a grassroots organization whose mission is to strengthen and spur native language processing research in African languages.40
The goal of advancing human intelligence for every person is timely, as tech companies and investors are poised to spend billions on developing artificial intelligence. Most of the evaluations currently used to measure gen AI focus on problem solving, reasoning, and knowledge across academic disciplines.34 However, some also assess qualities such as meeting community-driven human preferences or being harmless to people.35 Taking these into account, evaluations for gen AI can be concordant with SDG targets for quality education, measuring success in what the effects of a quality education are intended (or hoped) to be. Said another way, both the gen AI companies focused in improving model intelligence and those people who are working on quality education from the UN SDG perspective should be aware of what each other are up to and work for a convergence of metrics that reflect the objectives of both and, particularly, can help improve the quality of education in Africa.
Africa needs gen AI apps available for age-appropriate use by every person and in every language of Africa and ways to engage families, schools, governments, and other institutions in work with gen AI as a partner. Major companies developing gen AI—including Google, Microsoft, and Anthropic—are already sponsoring educational projects using it in Africa.41 These efforts must be expanded and ramped up with support from governments at all levels.
Gen AI tech leaders are bullish on intelligence, to say the least. Sam Altman, chief executive officer of OpenAI, stated this year, “Intelligence too cheap to meter is well within grasp.”36 If he is right, the nature of education could change dramatically, with a personal tutor for every student and access to all knowledge of the world, in every discipline, delivered in every language.
Providing abundance
Gen AI educational apps have already been introduced in Africa. For example, Rising Academies, founded in Sierra Leone, introduced a math app for students and an app for teachers using Anthropic’s Claude app, reportedly reaching one thousand schools across Africa and more than 150,000
“No poverty” is the first of the seventeen SDG goals, with seven listed targets.42 The first target is concrete: “By 2030, eradicate extreme poverty for all people everywhere, currently measured as people living on less than $2.15 a day.” Other
34.
See, for example: “Massive Multitask Language Understanding (MMLU) on HELM,” Center for Research on Foundation Models, Stanford University, last visited July 18, 2025, https://crfm.stanford.edu/helm/mmlu/latest/.
35.
“LMSYS Chatbot Arena: Live and Community-Driven LLM Evaluation,” LMSYS.org, March 1, 2024, https://lmsys.org/blog/2024-0301-policy; “Constitutional AI: Harmlessness from AI Feedback,” Anthropic, December 15, 2022, https://www.anthropic.com/news/ constitutional-ai-harmlessness-from-ai-feedback.
36.
Footnote 3, supra, Altman, “The Gentle Singularity.”
37.
“Rising Academies’ Chatbot Tutors Reach 150,000+ Students across Sub-Saharan Africa with Claude,” Anthropic, last visited July 18, 2025, https://www.anthropic.com/customers/rising-academies.
38.
These include the most advanced models of ChatGPT, Claude, and Gemini.
39.
“The Digital Transformation Strategy for Africa (2020–2030),” African Union, last visited July 18, 2025, https://au.int/sites/default/ files/documents/38507-doc-dts-english.pdf.
40.
Masakhane’s goal is for Africans to shape and own these technological advances. Masakhane’s website reports that its community consists of more than 1000 participants from 30 African countries and that it has provided translation results for over 38 African languages as well as a range of papers.
41.
Ben Gomes and Lisa Gevelber, “Growth Academy’s New AI Program to Help Startups Build Educational Tools,” Google, April 21, 2024, https://blog.google/outreach-initiatives/entrepreneurs/ai-for-education-startups/; “Opportunity International Works to End Farmer Poverty in Malawi with Azure AI,” Microsoft, March 24, 2025, https://www.microsoft.com/en/customers/story/23063opportunity-international-azure.
42.
“No Poverty,” United Nations Sustainable Development Goals, last visited August 26, 2025, https://www.un.org/ sustainabledevelopment/poverty/ .
ATLANTIC COUNCIL
5
Engaging generative artificial intelligence in African development
targets are more general; the last listed is: “Create sound policy frameworks at the national, regional and international levels, based on pro-poor and gender-sensitive development strategies, to support accelerated investment in poverty eradication actions.”
on LLMs, and digital connectivity has been central to many start-ups.47 Unsurprisingly, the McKinsey report is written with clients in mind and is fundamentally conservative. It doesn’t pitch a world of unlimited abundance for all with little need for human toil (and perhaps little need for advice from consulting companies). However, gen AI may be able to deliver much more than productivity improvement from speeding things up with little labor. Gen AI or AI is already being used to diagnose some maladies and prescribe treatment in African communities where highly skilled medical professionals aren’t present, although the practice raises issues of safety and accountability in the absence of human experts.48 Gen AI can forecast weather and assist in developing crops genetically engineered to resist harm from climate change.49 It can monitor and optimize diverse systems for agriculture, water supply, and energy.50
Gen AI leaders unabashedly assert that the technology will bring great abundance to Earth’s people. As Sam Altman said on his blog in June 2025: “In the 2030s, intelligence and energy—ideas, and the ability to make ideas happen—are going to become wildly abundant. These two have been the fundamental limiters on human progress for a long time; with abundant intelligence and energy (and good governance), we can theoretically have anything else.”43 Leaders of other firms have said similar things, and the book Abundance nicely describes a utopia that gen AI might provide, with less work and more holidays from increased productivity, and in which less work “has not meant less pay. AI is built on the collective knowledge of humanity, and so its profits are shared.”44
The McKinsey report also has a more stark but unsurprising omission: It does not spell out how the wealth from increased productivity would be shared with people whose jobs might be replaced by gen AI. How does the abundance get to them? Will the powerful and the wealthy, including gen AI leaders, keep the abundance for themselves?
A May 2025 McKinsey report on generative AI in Africa states that 40 percent of institutions in Africa have either started to experiment with gen AI or have already implemented significant solutions, with technology, telecommunications, and financial services having the highest levels of AI and gen AI adoption. The report says, “African economies could unlock up to $100 billion in annual economic value across multiple sectors from gen AI alone.”45 It provides estimates of how gen AI’s productivity will impact economic value across twenty sectors, led by retail; telecommunications; consumer packaged goods; mining, heavy industry, and energy; banking; and travel, transport, and logistics.46 The report also gives examples of gen AI initiatives in specific sectors. Africa has many such initiatives including both gen AI and AI not based
Lost employment is a significant and immediate potential adverse impact of using gen AI. With all the world’s information at its neural network fingertips, gen AI can do some work much faster than a human being. The jobs AI-enabled agents may be able to do include narrow tasks that existing gen AI models can already do with varying degrees of skill.51 A recent study from Stanford found that since the widespread adoption of gen AI, early-career workers in the United States (ages 22 to 25) in the most AI-exposed occupations have experienced a 13 percent relative decline in employment concentrated in occupations where AI is more likely to automate, rather than augment,
43.
Footnote 3, supra, Altman, “The Gentle Singularity.”
44.
Footnote 4, supra, Klein and Thompson, “Abundance.”
45.
“Leading, Not Lagging: Africa’s Gen AI Opportunity,” McKinsey & Company, May 12, 2025, https://www.mckinsey.com/capabilities/ quantumblack/our-insights/leading-not-lagging-africas-gen-ai-opportunity.
46.
Ibid.
47.
Aubrey Hruby, “Critical Connectivity: Reducing the Price of Data in African Markets,” Atlantic Council, March 3, 2023, https://www. atlanticcouncil.org/in-depth-research-reports/report/critical-connectivity-reducing-the-price-of-data-in-african-markets/.
48.
Rose Nakasi et al., “A dataset of blood slide images.” 2025. Science Direct, here.
49.
See, for example: “Nowcasting on Search Is Bringing AI-Powered Weather Forecasts to People across Africa,” Google, March 28, 2025, https://blog.google/intl/en-africa/products/explore-get-answers/nowcasting-on-search-is-bringing-ai-powered-weatherforecasts-to-users-across-africa/.
50.
See, for example: “Smart Grids and Predictive Maintenance: Africa’s Power Play with AI,” iAfrica, May 13, 2025, https://iafrica.com/ smart-grids-and-predictive-maintenance-africas-power-play-with-ai/.
51.
Narrow tasks include preparing, reviewing, and approving contracts, invoices, and expense reports; arranging meetings and travel; taking and summarizing notes of meetings; and writing code.
ATLANTIC COUNCIL
6
Engaging generative artificial intelligence in African development
human labor.52 A similar effect in Africa would be no surprise. AGI, if achieved, could be able to do intellectual work that is as challenging as anything a person might be asked to address. And robots are here to do physical work, with more on the way.
as a question of sovereignty.56 Africa’s Cassava Technologies, for example, is advancing a $500 million equity and debt plan for data centers across Africa with various investors.57 However, many questions remain. Are nations in Africa able to accommodate the exponential growth in data center energy that gen AI companies are seeking? Does it make sense for African nations to prioritize data centers at this time, given restrictions on the best GPU chips and trade-offs in other energy use and siting concerns?58 These issues share place with recent wellpublicized concerns that gen AI firms are overvalued and might collapse. Data centers to support the work could fail with them and leave African investors on the hook for debt. These are tough questions with no simple answers.
People won’t get to easy street if they lose their job and income. That inescapable concern has led to discussion about adopting policies such as universal basic income or other mechanisms to ensure guaranteed income.53 However, new jobs replacing those headed for extinction are not in the pipeline and, so far, the wealth from gen AI has gone to the investors and employees of the companies developing it. Now that gen AI hovers over us and potential job loss is imminent, it is past time for governments to get moving on the problem. Some form of tax transferring wealth from the richest few to the many poor is an apparent primary solution. The tax might wisely be placed on corporate income from the use of gen AI, asking the present and future gen AI billionaires to walk their talk of abundance—and so far, mechanisms for wealth transfer suggested by them are all talk.
Conclusion Generative AI needs to be regulated with both the strengths and weaknesses noted above in mind and a healthy dose of skepticism for corporate advocates, but ignoring the obvious value and use of gen AI does not make sense. Those concerned with development in Africa must engage with the technology and should consider its utility for reducing poverty and strengthening education, along with many other priorities such as digitizing and preserving languages. Gen AI has dangers and needs safety rails, especially for young people, but if it is not engaged then those youths and others whom it could harm without controls will be left behind. Not engaging with gen AI would be not only harmful but patronizing. More conversation is needed between those inventing and implementing gen AI models and those who work in development assistance, including those who helped make and are working toward the UN sustainable development goals. Two of these SDGs – no poverty and quality education – parallel the gen AI expectation—or boast—of future “abundance” and human or super-human intelligence. The SDG and gen AI camps need to explore what each has to offer for the other. It’s a two-way street.
AI data centers in Africa Data centers for gen AI computing power, and the energy they need, present another challenge to consider. Leading companies and some national governments, including the US government, are committed to serving the enormous energy appetites of gen AI data centers, minimizing competing energy needs and review processes for siting facilities and power transport connections.54 Estimates of new power needs for AI data centers globally range as high as 327 gigawatts (GW) by 2030, more than a 30-fold increase from about 10 GW in 2025.55 African business and governments are understandably interested in having data centers capable of developing and maintaining the most advanced gen AI models. Without that, they will depend on centers in wealthier nations. Some see it 52.
Canaries in the Coal Mine? Six facts about the recent employment effects of artificial intelligence. August 2025. Stanford University website, https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/ .
53.
Stanford Basic Income Lab website, https://basicincome.stanford.edu/about/what-is-ubi/ .
54.
Footnote 12, Supra, in respect to the US [Winning the Race: America’s AI Action Plan. July 2025. White House website.] and, for example, plans for new data centers in the United Arab Emirates.
55.
Konstantin F. Pilz, Yusuf Mahmood, and Lennart Heim, “AI’s Power Requirements under Exponential Growth,” RAND, January 28, 2025, https://www.rand.org/pubs/research_reports/RRA3572-1.html .
56.
Adam Satariano and Paul Mozur, “The Global A.I. Divide: Where A.I. Data Centers Are Located,” New York Times, June 21, 2025, https://www.nytimes.com/interactive/2025/06/23/technology/ai-computing-global-divide.html.
57.
Sebastian Moss, “African Tech Co. Cassava Technologies Secures $310m from Investors Including Google,” Data Center Dynamics, December 18, 2024, https://www.datacenterdynamics.com/en/news/african-tech-co-cassava-technologies-secures-310m-frominvestors-including-google/; “About,” Africa DataCentres, last visited July 18, 2025, https://www.africadatacentres.com/africa-datacentres-uses-strategic-investment-from-us-international-development-finance-corporation-to-expand-operations.
58.
For a wary perspective on what big gen AI firms are doing, see: Karen Hao, “Silicon Valley Is at an Inflection Point,” New York Times, May 30, 2025, https://www.nytimes.com/2025/05/30/opinion/silicon-valley-ai-empire.html.
ATLANTIC COUNCIL
7
Engaging generative artificial intelligence in African development
About the author William Yancey Brown is a nonresident senior fellow at the Atlantic Council Global Energy Center. He has held leadership positions in government, the nonprofit sector, and industry. Among these, he was the science advisor to Secretary of the Interior Bruce Babbitt in the Clinton administration and president and chief executive officer of the Bishop Museum in Hawaii, the Academy of Natural Sciences in Philadelphia, and the Woods Hole Research Center in Massachusetts.
About the center Led by Rama Yade, previously an ambassador and minister in the French government, the Atlantic Council’s Africa Center champions a new African narrative prioritizing investment opportunities and forward-looking partnerships. The center promotes dynamic geopolitical partnerships with African states, and the redirection of US and European policy priorities toward strengthening security and bolstering economic growth and prosperity on the continent.
ATLANTIC COUNCIL
8