Scaffolding for the South Africa National AI Policy Framework
Scott Timcke, Zara Schroeder, Drew Haller / Nov 29, 2024Research ICT Africa's Scott Timcke, Zara Schroeder, and Drew Haller write that South Africa's 2025 AI Framework aims to harness AI for public value and development while managing risks but needs clearer implementation and integration plans.
In the third quarter of 2025, the South African government released its National AI Policy Framework. Officials intend for the Framework to be the first step of a process that culminates in a National AI Policy or an AI Act. Given South Africa’s major socio-economic challenges like persistent poverty, inequality, and unemployment, their hope is that regulatory action can position AI as a driver of public value creation and well-being. While not viewed as a panacea, officials see AI as a potential source of economic growth and development.
The Framework rests on nine strategic pillars. These pillars—talent development, digital infrastructure, research development and innovation, and fairness and mitigating bias, for example—seek to position South Africa at the global forefront of responsible AI advancement, an aspiration that South African officials are not shy of pronouncing. The South African vision is to focus on risk management for human-centered AI, by which they mean that AI technologies be designed to complement human decision-making and enhance human capabilities, rather than replace them.
Creating conditions for downstream success
The Framework’s emphasis on evidence-based policymaking for the public interest stands out as a particular strength. This recognition that sound AI policies must be grounded in empirical data and measurable outcomes demonstrates a commitment to practical, results-driven governance rather than speculation. In this spirit, there is space for refinement.
The lack of specific responsibility assignment and cross-sectoral coordination mechanisms undermines the framework’s utility in guiding downstream activity. It is not too early to start articulating appropriate institutional arrangements, or encouraging debates between different models. A proposed multi-stakeholder platform to guide implementation lacks details about representation, participation criteria, and decision-making processes. This institutional uncertainty is further complicated by strained budgets and unclear funding mechanisms for new structures.
Next, the framework’s lack of integration with existing policy landscapes is inadequate. There is a value in horizontal policy coherence across trade, competition, and other sectors. Reference to South Africa’s developmental policy course as articulated in the various Medium-Term Strategic Frameworks and in the National Development Plan 2030 would be helpful. There is a focus on transformation, development, and capacity-building, strengthening the intentions set out in the 2019 White Paper on Science, Technology and Innovation, which emphasizes ICT's role in further developmental goals within a socio-economic context that features high unemployment rates.
In the evolving landscape of global AI governance, the Framework could account for the inherently transversal nature of AI, where ineffective national strategies can cause significant ripple effects across interconnected digital ecosystems. Accordingly, South Africa could align its AI governance with international standards like the UN Global Digital Compact, UNESCO’s recommendations on the ethics of AI, and the final report from the UN High-level Advisory Body on AI.
Closer to home is the African Union Development Agency’s (AUDA-NEPAD) AI white paper and the AU’s Continental Artificial Intelligence Strategy. (Abdullahi Tsanni has an essay in MIT Technology Review on the politics between these initiatives and their blocs of funders.) The main point is if South Africa wishes to be a global player in AI governance, it needs to confirm its active participation in international AI policy forums, and show the utility of these venues to inform national agendas.
Avoiding structural weaknesses
Research ICT Africa’s After Access surveys reveal profound demand-side barriers that fundamentally challenge meaningful digital connectivity and, by extension, widespread AI adoption. While the Framework acknowledges the necessity of expanded internet coverage, it neglects accessibility factors, with affordability being one most significant obstacle to AI uptake and utilization. To ensure truly inclusive AI development, the Framework must directly confront the economic challenges preventing digital access, particularly in underserved and rural communities. Bridging the digital divide requires a comprehensive strategy that transforms connectivity from a privilege to a fundamental right, enabling equitable participation in the emerging AI ecosystem.
The treatment of the sources of discrimination and economic inequalities in the Framework requires conceptual strengthening. While the framework discusses economic maldistribution, it lacks mechanisms for ensuring equitable access to AI opportunities. On this note, the Framework could do with a thorough consideration of how to prevent human exploitation, beyond just the introduction of human oversight in AI development. This consideration could extend to the impact of AI impact on labour’s bargaining power and overall economic positioning.
The profound digital divide demands a strategic approach to ensuring equitable access to AI technologies, especially in underserved and rural communities. These disparities require a targeted framework that can leverage AI as a transformative tool for addressing historical economic and social challenges. The current Framework should articulate a clear vision for how AI can be deployed to bridge technological and economic gaps, ultimately serving as a catalyst for more inclusive development.
The Framework requires more attention to market dynamics and economic implications. It inadequately addresses the concentration of AI capabilities within specific sectors, where certain technology firms gain disproportionate advantages through superior data access. The Framework needs clearer mechanisms for incentivizing AI adoption among startups across various sectors while preventing market monopolization.
While acknowledging the transformative potential of AI, the Framework falls short of addressing geopolitical considerations and global inequalities. Changes in the international system, including major powers’ efforts to repatriate production and potential conflicts like those surrounding Taiwan, will significantly influence technological development. The Framework omits key economic risks, particularly regarding supply chains and production.
Moreover, the Framework overlooks the concentrated control of hardware and infrastructure by firms primarily based in the US, Europe, and China. For countries like South Africa, developing strategic long-term access to hardware through targeted trade pacts and agreements is crucial. State planners need a more comprehensive analysis that contextualizes AI development within the current conditions.
Several key areas in the Framework could be improved with an additional layer of explanatory detail. For example, it would be instructive for officials to recognize that AI is not a single technology but rather a diverse array of advanced models. Each of these models in turn may require different regulatory approaches. Regulatory approaches that view AI as a ‘black box’ might not maximize the intended benefits.
Enabling public value creation
Few would disagree that South Africa’s extensive socio-economic inequalities represent significant obstacles to AI adoption. This is one reason why the Framework discusses public value. Unlike the private sector’s focus on profit maximisation, public value creation focuses on improving access to essential services, promoting social equity, and addressing systemic inequalities. The Framework stresses public value creation, focusing on using AI healthcare, education, finance, agriculture and improved public services.
While risk management is a useful paradigm for governance – and can direct policy attention to labour market disruption and encoded bias – the Framework could be more imaginative in placing public value creation at the centre of the document. This articulation can help prioritize interventions and guide the development of appropriate safeguards to entrench and maintain the appropriate protections required to produce public value.
It is important that the strategy ensures that AI technologies are accessible to all South Africans, particularly disadvantaged communities. This requires investment in digital infrastructure, education, and skills training. In this respect, it is important to distinguish between an enabling environment for policy development and an enabling environment for public value creation. Central to this strategic positioning is the government’s critical role as a primary data collector in any digital public infrastructure and digitalization programs, which necessitates robust mechanisms for data protection, categorization, and minimization.
Towards a strategy
The Framework’s approach to talent development in the AI sector requires refinement. While it broadly advocates for incorporating AI into educational curricula, it fails to clearly differentiate between teaching AI literacy and using AI as a tool for teaching. This lack of clarity is concerning given the digital inequalities that exist in South Africa. Additionally, the Framework does not adequately address the pressing issue of talent retention, overlooking the risks of brain drain and the potential for digital labor extraction, which could undermine the country’s capacity to build a sustainable and competitive workforce.
To address these challenges, a more strategic approach is needed—one that focuses on strengthening foundational digital skills while also developing effective mechanisms to retain skilled professionals within the country. The Framework should set clear objectives and investment targets to guide talent development, ensuring that South Africa can not only train but also keep its AI talent.
A coherent public policy agenda can serve broad goals, like enhancing social welfare or encoding democratic practices. This approach can guide policymakers toward creating interventions that are both pragmatic and responsive to the needs of citizens.
Related Reading
Authors
