The OECD-African Union AI Dialogue 2.0: From strategy to implementation

The OECD-African Union (AU) AI Dialogue, held in Cairo from 19 to 21 November 2024, brought together participants from more than 30 countries—including 20 AU member states—to work together on concrete action steps to implement the Continental AI Strategy. As a follow-up to the first OECD-AU AI Dialogue in March, the event featured more than 40 speakers across 11 sessions focused on achieving the strategy’s key objectives.

The event was hosted by Egypt’s Ministry of Communications and Information Technology (MCIT) and supported by the United Kingdom (UK) Foreign, Commonwealth & Development Office (FCDO), with the programme shaped by the OECD. Dr. Conrad Tucker, Director of Carnegie Mellon University – Africa (CMU-Africa), chaired the event. The dialogue was kicked off with opening remarks by:

• Dr. Hoda Baraka, Advisor to the Minister of ICT for Technological Talents and National AI Lead, Egypt
• Juraj Čorba, Chair, OECD Working Party on AI Governance; co-chair, Global Partnership on AI (GPAI); Senior Expert on Digital Regulation & Governance; Slovak Republic
• Nicole Gregory, Head of the International Technology Department, UK FCDO.
• Ambassador Ashraf Swelam, Assistant Minister of Foreign Affairs for African Organisations, Egypt
• Amine Idriss Adoum, Director for Infrastructure, Energy, Industrialisation and Regional Integration; African Union Development Agency (on behalf of the African Union Commission).

The African context

As emphasised by Juraj Čorba in his opening remarks, it has been a momentous year in the field of AI. Models are demonstrating stronger capabilities than ever, and AI companies have reached an estimated USD 9 trillion in value. However, challenges such as automated bias and discrimination remain pressing concerns, and improving but still fragmented international governance leaves it challenging for governments to effectively address the transnational impacts of technologies that know no borders.

However, one thing remains unchanged. As discussed in a new OECD report published days before the dialogue, the development and use of AI and the resulting benefits are largely limited to advanced economies. Emerging and developing economies may be more exposed to the potential risks of AI, with developed economies leading the AI transformation better able to absorb change, the potential cutting off of traditional development pathways, and the exploitation of inexpensive labour to support AI advancements. Adding to this, governance and risk management approaches that seek to mitigate harms are largely informed by the contexts and values of advanced economies, leaving underrepresented populations further behind.

The July adoption of the Continental AI Strategy and African Digital Compact, which contributes to the UN’s Global Digital Compact, represent major feats that have the potential to position Africa as an integral participant in the AI revolution and to make its voice heard in global AI governance. While the strategy provides a strong foundation for transformation, implementation is inevitably the hardest part. The goal of the OECD-AU AI Dialogue was to help AU member states achieve this in an effective way that also pays tribute to Africa’s many contexts and realities.

Themes and lessons learned

Several key themes emerged during the event. First, there was discussion on the plethora of African AI technical and policy initiatives, led both by governments and by stakeholders in grass-roots initiatives. Africa is not starting strategy implementation from scratch, and regional cooperation is a must to bring together ongoing and future activities. Other themes included:

• Recognise and respect Africa’s diverse contexts, cultures, languages, and realities. Africa is a continent of diverse people, cultures, languages, and economic realities. Regional approaches are needed to reflect this while ensuring inclusion and mutual respect. While AI policies and governance frameworks from other parts of the world can inspire approaches in the region, they need to be contextualised through an African lens.
• AI is here to stay. It’s up to us those in the region to develop the relevant frameworks and tools to ensure that AI’s power is directed towards a greater good and that its risks are mitigated, which can feel like an uphill, but critical, battle in Africa. This includes stronger engagement in international AI governance beyond the region.
• “AI” is not synonymous with the biggest, most powerful frontier models. AI for Africa means addressing African needs and nourishing locally developed and driven solutions. This often means pursuing narrow, purpose-driven AI applications, which are localised and tailored to specific community needs, empower local actors, and tend to trigger less risk aversion among key stakeholders.
• Trust is key. The OECD Trust Survey shows that nearly half of people in OECD countries have no or low trust in government, with results noticeably lower for women, young people, and those with a sense of financial insecurity or who feel a lack of political voice. Surveys among African countries yield comparable results. African governments need to engage the public to build trust in governments’ abilities to manage the AI transformation on their behalf.
• Capacity building is critical. Regional actors have already developed relevant programmes to build skills and abilities among different target groups – children, women, civil servants – but more needs to be done to increase equal opportunities.
• Effective implementation. Strategies are important, but actions speak louder than words. AI actors in Africa need to get started with AI projects in a hands-on way, even if small, to really understand AI’s potential and to build capacities in the public administration and society.

By empowering local communities, integrating AI into public services, and fostering collaboration, Africa can unlock the true potential of AI and build a foundation for its long-term, sustainable success. African countries and digital leaders are confronted with the same global issues in seizing AI’s benefits while mitigating its risks, and they are the now taking the next steps in implementing the Continental AI Strategy.

Event highlights

Driving the implementation of the Continental AI Strategy

Two sessions bookended the event focused on high-level implementation insights, with the first session being a discussion on “Driving the implementation of the Continental AI Strategy“. Amr Safwat from Egypt’s MCIT kicked off the session with a presentation on the Continental Strategy’s contents, objectives, and next steps. He then moderated a panel discussion among speakers involved in its development. Amine Idriss Adoum touched on the need for systematic digital capacity building in the region and the potential for AI in Africa but also cautioned that governments need to resolve long-standing infrastructure challenges, including access to stable electricity. Rachel Adams, one of the lead drafters of the strategy, raised that although the strategy is regional, there are still disparities in digital maturity and access to AI enablers, stressing the need for further regional cooperation in areas such as policy interoperability, data governance and sharing, equitable distribution of benefits, and empowering women to pursue education in STEM fields. Ahmed Tantawy concluded the session by providing an overview of the efforts of Egypt’s Applied Innovation Centre (AIC). He provided a call to action for African states to go beyond strategy documents and develop hands-on AI projects to “drive adoption by doing things”.

The final session of the event featured remarks by Amr Talaat, Egypt’s Minister of Communications and Information Technology and others on insights gleaned from the OECD-AU AI Dialogue and next steps for strategy implementation, as further covered in articles from Daily News Egypt and TV BRICS.

Developing an African Charter on Trustworthy AI

An important part of implementing the Continental AI Strategy will be the development of an African Charter on Trustworthy AI, which will provide a framework for responsible and ethical AI tailored and contextualised for Africa. The dialogue featured an interactive workshop for participants to establish the initial contours of such a charter, with workshop design and lead facilitation by Angela Hanson from the OECD Observatory of Public Sector Innovation (OPSI).

There is no shortage of frameworks for ethical and trustworthy AI, with hundreds of such instruments existing around the world. However, African voices are underrepresented in the development of international instruments, and only a handful of African countries have articulated principles and frameworks tailored to their specific needs (e.g., Egypt, Nigeria, Rwanda, South Africa).

Based on OECD analysis of relevant frameworks and a taxonomy of 17 AI principles based on a synthesis of over 200 instruments by AI research and workshop participant Nicholas Kluge Corrêa, the workshop sought to:

• Establish common ground and shared understanding among participants of what it means to develop this charter in the African context.
• Consider the extent to which existing principles and approaches fit in an African context, and determine what elements may be missing that are important to recognise on the continent.
• Articulate how AI principles should be contextualised in a way that makes them uniquely African.

Several recurring themes came from the workshop discussions, including a few priority principles:

• The value of diversity: The importance of contextualising AI to reflect Africa’s diverse cultures, values, languages, and levels of development in digital infrastructure while also addressing the continent’s unique challenges and reflecting priority sectors.
• Inclusion: Ensuring that Africa’s approach promotes broad accessibility of AI inputs, tools, and applications and is mindful to leave no one and no area behind.
• Human capacity: How Africa can leverage its growing population and build skills with AI to create value at home.
• Accountability: Developers and deployers of AI technologies should be compliant with regulatory bodies and be accountable for their actions and the impacts caused by their technologies, especially in the Global South.

The workshop resulted in many insights and considerations, and AU delegates plan to leverage the workshop results as the first step in articulating the initial draft of a full charter.

Deep dives into critical sectors

The all-women panel on “Promoting responsible AI in public service delivery” included a variety of perspectives ranging from local to international. Teddy Nalubega, coming from rural Uganda, discussed the need to develop AI-enabled public services that are tailored to community needs, as touched on in her recent report. She and Hoda Baraka emphasised capacity building among the region’s youth to use and understand such services, given the young demographics of Africa, as well as public servants. Dr. Baraka also reflected on the Egypt AIC and the potential for AI to improve service delivery, emphasising the need for hands-on implementation and enforcement of AI strategies and policies. Sarah Handyside and Angela Hanson underscored the importance of citizen trust in governments’ ability to manage AI as critical, detailing the UK’s Algorithmic Transparency Recording Standard (ATRS – OECD case study) and the OECD trust survey, respectively. Ms Hanson also discussed the findings from the recent OECD report “Governing with AI: Are governments ready?“. Finally, Leila Elmergawi touched on the potential of tailored, purpose-driven AI services for Africa (e.g., early warning systems tools to optimise the electrification of rural areas). She also connected the discussion with the US AI Executive Order and Partnership for Global Inclusivity on AI (PGIAI), which helps resource AI efforts with compute, capacity building, and local datasets.

In another domain-specific session on “AI in agriculture“, speakers discussed ways to leverage AI to enhance food production. In a panel moderated by Dumisani Sondlo, Erik van Ingen outlined the Food and Agriculture Organisation’s (FAO) approach to responsible AI, including a walkthrough of resources and real-world applications. Kathleen Ziemann also provided international perspectives, presenting on the GIZ FAIR Forward initiative to promote open and sustainable development and application of AI, including for agriculture, in Africa and Asia. Eli Pollak provided perspectives from industry and unpacked the case of the AI-driven regional agricultural financing company Apollo Agriculture. Finally, Maha Gmira discussed how the creation of centres of excellence can support the use of AI in agriculture, and how generative AI in this field can create a synergy between AU countries.

Exploring and mitigating AI risks

Three sessions touched on better understanding AI risks and what can be done to manage them. In his keynote remarks, Seydina Moussa Ndiaye spoke about the efforts of the UN High-Level Advisory Body on AI, on which he served, and presented the findings of their “Governing AI for Humanity” final report. This report canvasses key AI risks and makes a series of recommendations for governance and policy actions to mitigate them.

Later in the agenda, Dr Ndiaye joined a panel discussion on “Exploring and mitigating AI risks“, which was structured around three sub-topics:

• Identifying risks. Dr. Ndiaye dug deeper into the risks uncovered by the AI Risk Global Pulse Check global expert crowdsourcing exercise, which identified key AI harms of concern (results in UN report p. 88). Jamie Berryhill and Stijn Broecke presented on the OECD AI Incidents Monitor (AIM) and potential risks to labour markets, respectively.
• Understanding/evaluating risks. Amit Thapar, who also moderated the session, presented on the UK AI Safety Institute (AISI) and its work to date on the International Scientific Report on the Safety of Advanced AI. Merry Walker provided perspectives from the US AISI, as well as efforts to build an international network of AISIs.
• Mitigating risks. Mr. Berryhill discussed 10 priority policy recommendations discussed in the new OECD report, “Assessing potential future artificial intelligence risks, benefits and policy imperatives“.

Another relevant session focused on “Promoting AI experimentation and building AI sandboxes” and key approaches to managing risks in a controlled way. Moderated by Rosine Urujeni, the session featured a discussion on a variety of regional, national, and non-governmental sandbox efforts from outside the region to help AU delegates learn best practices to inform the development of their own mechanisms. Alex Moltzau presented the EU’s approach, which includes a requirement for each EU Member State to develop an AI sandbox by 2026, with Marie-Léa Rols discussing France’s efforts to do so. Joseph Phillips presented the UK’s experiences, lessons learned and future plans, such as developing a regulatory sandbox for AI medical devices. Armando Guio walked participants through a methodology for exploring sandboxes in developing economies. Abby Gilbert discussed the Institute for the Future of Work’s (IFOW) Responsible AI Sandbox for promoting trustworthy AI applications in the workplace.

African engagement in multi-stakeholder communities and international AI governance

A discussion on “Implementing Africa’s AI vision: Insights and engagement with industry, academia and civil society” considered perspectives from non-governmental actors. Jacques Ludik provided private sector insights and discussed Africa’s demonstrated ability to leapfrog others with technological innovation, as it did with mobile payment services. He stated that democratising AI is critical and will require strong digital and physical infrastructure, capacity building, and venture capital investments. Representing academia, Nagla Rizk, one of the developers of Egypt’s national AI strategy, discussed the efforts and results of the MENA Observatory on Responsible AI at the American University in Cairo. She stressed the need for open stakeholder engagement and the inclusion of women in AI. Representing civil society, Nyalleng Moorosi discussed the importance of inclusive datasets and inclusiveness in designing AI systems, as “the types of projects we want to explore are defined by who we are and what resources we have”. She cautioned against some types of automated decisions (e.g., benefits calculations) and encouraged the development of accountable, purpose-driven small models that can work in low-resource environments. She and Florence Nakazibwe believe there is momentum growing in the AI ecosystems among Africans, both in the region and part of the diaspora, and are working to forge connections and consortia to further this momentum. Beyond the common refrain of focusing on AI risks, Ms Nakazibwe also pushed for a novel consideration for AI to be recognised as an enabler for human rights and advocated for new research on aligning AI systems with human rights. Tabani Moyo pointed out that while a continental strategy is helpful, individual countries need to develop their own plans that take into account their economic realities and relevant non-governmental AI actors.

The session on “Strengthening African engagement in multilateral initiatives” featured speakers from both within and outside the region. From outside the region but with an aim for enhancing African engagement, Juraj Čorba presented on the new integrated partnership between the OECD and GPAI, which brings together OECD Members and GPAI countries – 44 in total – and aims to welcome new members, including developing and emerging economies. Mr. Čorba explained that more information on GPAI membership will be made available soon and, in the meantime, welcomed more African participation in the OECD GPAI Expert Community. Hanne Juncher spoke about the Council of Europe’s (CoE) Framework Convention on AI and Human Rights, Democracy and the Rule of Law, the first legally binding international treaty on AI, and how AU members could sign on and pursue implementation through the HUDERIA risk assessment process.

From inside the region, Ambassador Amr Aljowaily discussed the importance of African participation in all international AI governance forums, as covered in an article by See News. Angela Casell Bush highlighted Liberia’s commitment to participating in the implementation of AI across the African continent and engaging in international collaborative efforts, as further reported by the Liberia Telecommunications Authority.

Immersion in Africa’s technological excellence, both past and present

The event featured a day dedicated to site visits that exemplify Africa’s technological accomplishments. In learning about the present (and future) of technological capacities in the region, event participants had an opportunity to tour and hear from the experts at the Technology Innovation & Entrepreneurship Center (TIEC), a driver of entrepreneurial innovation in Egypt. Participants learned about the TIEC’s work in robotics, gaming, the Internet of Things (IoT), and digital capacity building. In focusing on the past, no event in Cairo would be completed without seeing the Egyptian pyramids of Giza, built 4,500 years ago. The site visits offered tangible connections with the themes and aspirations discussed throughout the dialogue, which further inspired new ideas and fostered further connections among delegates and participants.

The path ahead

This second dialogue comes at a very timely moment, with several developments taking place at the international level, including the creation of the OECD-GPAI Integrated Partnership, the adoption of the UN Digital Global Compact, and the upcoming convening of the AI Action Summit in France.

As the dialogue concluded, it was clear that this event was not an endpoint but a milestone in further AI technological and governance advancements within the region and strengthened African engagement in AI governance globally.

The coming months are sure to be eventful for AU member countries and their efforts on AI. With the implementation of the Continental AI Strategy in full swing and the development of an African Charter on Trustworthy AI officially underway, we are excited to see the event’s delegates take action to seize the promise of AI for good in Africa.