Seven months after the African Union Executive Council approved a conceptual framework for Artificial Intelligence (AI) on the continent, the AU has adopted the continent’s strategy for AI.
This strategy, adopted in July, marks a significant milestone in Africa’s journey toward AI integration and development.
The origins of this strategy date back to the Third Ordinary Session of the Specialised Technical Committees on Communication and ICT (STC-CICT) held in 2019 in Egypt. During this session, the continent’s first working group on AI was established with a threefold mandate: to create a unified stance on AI, design a continental framework, and establish an AI think tank.
This think tank was tasked with making recommendations on how AI could help achieve the Sustainable Development Goals (SDGs) and Africa’s Agenda 2063.
Adaptation to African Realities
The African Union Commissioner for Infrastructure and Energy, Dr Amani Abou-Zeid, who led the design of the strategy document, noted that the adaptation of AI to African realities – vis-a-vis the creation of competitive African AI markets is critical, and the goal is to ensure AI systems are designed to accommodate the diversity and contextual realities of Africans.
Overall, Abou-Zeid believes that the continental strategy will provide a common vision and path to accelerate responsible AI innovation and adoption in Africa. But how does the strategy intend to make this happen? And to what extent are these feasible? Here are a few takeaways:
Governance and Regulation: A Non-Negotiable Aspect
AI governance and regulation are non-negotiable. A major part of the continental strategy emphasises the need for mechanisms to govern the design and use of AI systems to ensure they are accountable to people across the continent.
The need for a strong governance and regulation regime is considered non-negotiable because even when designed for purely legitimate reasons, AI systems can be wielded to harm people.
But the AU strategy document does not only want a governance regime, the strategy also calls for a multi-tiered governance approach involving a variety of stakeholders.
Action items include amending and applying existing legal frameworks, closing regulatory gaps that have allowed unacceptable practices like labour exploitation for AI by countries like OpenAI in Kenya, and developing ethical standards for AI within the context of African realities.
Similarly, the continental strategy considered the development of national policy frameworks for AI in African countries a necessity. This is not far-fetched. Currently, only six out of Africa’s 54 countries – Algeria, Benin, Egypt, Mauritius, Rwanda and Senegal have developed stand-alone AI strategies.
This pales comparatively to what is obtainable globally because, as the document highlights, between 2017 and 2023 alone, 67 countries developed national AI strategies.
National AI strategy documents are important because they provide a framework for AI development in a country and the direction for its application.
To expand the currently limited number of countries, the AU strategy recommends that the African Union support member states in developing national AI policies by integrating them into national development plans. peer learning, readiness assessments, and knowledge exchanges are all proposed as ways to drive this forward.
Data Quality and Accessibility
The continental strategy also recognized the need for improvement of datasets available for AI development, including how these data are accessed. AI models are built on a massive amount of machine-readable data.
However, African countries are home to a significant volume of dirty data, simply put, incomplete, inaccurate, and inconsistent data that are not effective for developing AI systems, and a significant proportion of the available one is not easily accessible.
To correct this, the strategy document recommends the creation of more openly available datasets that underpin AI development. But more importantly, it emphasised the need for African countries to invest in the creation of high computational processing power and data resources for AI development.
These resources are critical infrastructures necessary to store and process the volume of datasets that can power AI development but currently, less than two per cent of data centres with this capacity are in Africa.
Data Protection Framework
In the same vein, the document provided a framework for data protection, one of the most critical issues relating to AI development globally.
Currently, about two-thirds of African countries have data protection policies in place, but the continental strategy recommends the need for these policies to address critical issues like localisation, data classification and cross-border data transfers in line with the recommendations of the AU Data Policy Framework which itemises standards for sharing data ethically, responsibly and securely.
This recommendation is not out of place because in countries like Nigeria where a data protection framework exists, there are legal and technical loopholes that undermine their effectiveness.
AI in Public and Priority Sectors
The strategy document also recommended the use of AI in public and priority sectors. That AI can boost the effectiveness and quality of service delivery by a country’s public sector is not exactly new – the European Union alone has over 600 examples of AI usage across its public sector.
However, this level of usage is currently limited in Africa and the AU continental strategy wants to change this. To achieve this, the document recommends that member states should encourage the development of digital government strategies that integrate AI and enhance capacity building among civil servants.
However, the strategy document did not outline how African countries should effectively do this within the context of Africa’s current realities like access to electricity supply and broadband network which are still big issues for policymakers, without alienating citizens at the lower rung of the social ladder.
For priority sectors, the strategy document recommends the adoption of AI in areas like Agriculture, health, education, and climate change adaptation.
It made a wide range of recommendations that will enhance the capacity of member states to increase productivity and enhance efficiency in these sectors.
However, it is important to note that the effectiveness of the application of AI in these priority sectors is reliant on several critical issues like data protection, and infrastructural improvements.
Building AI Research Capacity in Africa
Also, it made a case for the enhancement of the capacity for AI research in Africa. Even though the contribution of Africans to research on AI is increasing – for example, an analysis of contributions to Github shows that the percentage of actual total contributions from African authors has increased by two per cent between 2010 and 2020, overall the continent has contributed just 2.3 per cent of the available research.
This must change for several reasons: to develop AI solutions and viewpoints in line with Africa’s realities, to ensure the adequate representation of African language and culture by AI systems, and to create AI timelines in line with the development agenda of African countries, among others.
The continental strategy recognises all these and recommends the investment of financial and technical resources necessary to fill the research gap on AI across the continent. While this is necessary, the strategy did not speak to the other component of this: where would these financial resources come from?
Governments in African countries like Nigeria that have made increased efforts towards AI development face criticism from citizens who believe there are more fundamental development issues that this funding should be allocated, and this scepticism is shared by more people across Eastern, Central and Western Africa than anywhere else globally.
Implementation Timeline and Challenges
Finally, the continental strategy has a five-year implementation timeframe with two phases. The first phase between 2025 and 2026, will focus on areas like governance frameworks, resource mobilisation, and capacity building, and the second phase will focus on the implementation of core projects and actions of the strategy.
While the strategy outlines a monitoring and evaluation component that includes the introduction of an AI readiness index for Africa and a midterm progress review, the implementation component is perhaps its weakest aspect because at best, it remains an advisory document with little or no incentive to ensure member states effectively work towards its implementation.
What this means is that for the recommendations of the strategy to be effective, an implementation roadmap must be outlined as an addendum to it in the coming months.