Spiced Paper

A successful first Global Dialogue on AI Governance would produce more than broad agreement on the importance of responsible AI. It should create a clear, practical foundation for how AI can be used to advance human development without undermining human dignity, employment, economic stability or national sovereignty. One key outcome should be a shared understanding of where AI should be used, where it should be limited, and where human authority must remain non-negotiable. This is especially important in areas that affect people's livelihoods, rights, access to services, education, health, justice and participation in society. The Dialogue should also address the economic impact of AI, particularly for developing countries and countries with high unemployment. AI should not be adopted only for efficiency or cost reduction. It should be used to build capability, strengthen public infrastructure, improve productivity, support new forms of decent work and expand opportunity. Countries should be supported to make sovereign decisions about AI adoption in line with their labour markets, development priorities and social realities. A further measure of success would be practical cooperation between countries, including support for AI capacity-building, digital infrastructure, skills development, data governance and locally relevant AI systems. Global governance should not deepen dependency on a small group of powerful countries or companies. Finally, the Dialogue should establish a pathway toward accountability: clear governance standards, risk assessments, transparency expectations, human oversight, and consequences for harmful or irresponsible AI deployment. The purpose of AI should be the evolution of humanity, not the disruption or weakening of people, communities and economies.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Spiced Paper selected these priorities because responsible AI governance must begin with human understanding, capability and protection. AI capacity-building is urgent because many individuals, workers and organisations are already using AI without fully understanding its benefits, risks, limitations, biases, data implications or possible long-term effects. African countries need accessible AI education at a basic and practical level, including public learning, workplace training, and support for organisations whether they choose to adopt AI immediately or not. The social, economic, ethical, cultural, linguistic and technical implications of AI are equally important. In multilingual and diverse societies such as South Africa and across the African continent, AI systems must reflect local languages, realities, labour markets, cultural contexts and development needs. AI should strengthen people's ability to earn, improve dignity in work, and support economic participation rather than create fear, exclusion or unnecessary displacement. The protection and promotion of human rights is central because people are better protected when they understand how AI affects them and when human agency remains clear. Individuals should know when AI is being used, how it affects decisions, and when human judgement must remain final. Transparency, accountability and human oversight are necessary across the AI value chain. Developers, organisations and deploying institutions should be clear about how AI systems are built, used, monitored and governed. Accountability should not be shifted entirely to end users when systems are designed, procured or mandated by others. Overall, these priorities support AI use that advances education, prosperity, economic development and human dignity.

Yes. One cross-cutting issue that should be more explicitly recognised is the direct impact of AI on the human being: human agency, cognition, behaviour, connection, confidence, dignity and the ability to earn. The listed themes are important, and many of them do touch human impact indirectly. However, because AI is ultimately being introduced into human life, work, learning, decision-making and social systems, the effect on people should be treated as a distinct area of inquiry and engagement, not only as a secondary consideration within technical, economic or governance themes. AI can support human development, access to knowledge, creativity, productivity and inclusion. At the same time, it can also affect how people think, make decisions, trust themselves, relate to others, participate in work, experience fear, retain skills and understand their own value. These effects are not only technical or economic. They are deeply human. For this reason, a dedicated focus on human agency and human impact would create space for broader public participation, interdisciplinary research and lived-experience feedback from workers, educators, young people, communities, employers and civil society. It would help ensure that AI governance is not only system-centred, but genuinely human-centred. The success of global AI governance should therefore be measured not only by how well AI systems are regulated, but by whether AI strengthens human beings: their ability to think, choose, work, earn, connect, learn and participate meaningfully in society.

In South Africa and the wider African context, governance gaps in AI are creating both significant risks and important opportunities. A key challenge is the gap between AI use and AI understanding. Many individuals, workers and organisations are already using AI tools without sufficient knowledge of data privacy, bias, hallucinations, verification, human oversight, cognitive dependency or the long-term effects on work and decision-making. This creates risk for individuals, employers and public institutions. Another major challenge is cultural, linguistic and contextual relevance. Many AI systems are trained largely on data that does not fully reflect African societies, languages, accents, labour markets, histories or lived realities. This can result in inaccurate, biased or incomplete outputs, especially where AI is used in education, recruitment, public services, policy, research or business decision-making. There is also a trust and policy gap. South Africa's recent withdrawal of its draft AI policy, following concerns about unverified or fictitious sources, shows why human oversight, transparency and credible governance processes are essential. Public confidence will be weakened if AI governance is not carefully explained, verified and communicated. AI also raises serious economic concerns in a country with high unemployment and inequality. Without clear leadership, people may experience AI as a threat to jobs, income and dignity. However, if governed well, AI can support reskilling, entrepreneurship, service delivery, productivity and new forms of decent work. Finally, AI sovereignty must be linked to infrastructure reality. Data centres, energy supply, connectivity and digital access are critical. For Africa, the opportunity is to build AI capability in a way that strengthens local economies, protects human rights and ensures people are not only users of AI, but active participants in shaping it.

AI Dialogue can play an important role in advancing international cooperation by moving global AI governance from general principles to practical commitments that support all countries, not only those already leading AI development. For developing countries, including many African countries, the challenge is not a lack of intelligence or ambition. It is often a lack of infrastructure, funding, compute access, reliable energy, local data capacity and meaningful participation in the design of AI systems. International dialogue should therefore create pathways for cooperation on AI infrastructure, sustainable data centres, digital public infrastructure, technical support, local language models, responsible data governance and skills development. The Dialogue can also help ensure that digital sovereignty does not become isolation. Countries such as South Africa should be able to make sovereign decisions about AI use while still participating in global innovation. This requires interoperable governance approaches, where countries can align on safety, transparency, accountability and human rights, while adapting AI governance to their own labour markets, cultures, languages and development priorities. A successful Dialogue should also bring technology developers, governments, civil society, educators, workers and communities into the same conversation. This is important because AI governance cannot be shaped only by countries or companies with the most advanced systems. Those affected by AI must also help define how it is used. International cooperation should include fair access to AI capability, shared research, transparent standards, funding mechanisms, public-private partnerships and support for locally relevant AI systems. The goal should be to ensure that AI strengthens human development, economic participation and national capability, rather than deepening global inequality or technological dependency.

The AI Dialogue should build upon existing initiatives such as the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, the Global Partnership on Artificial Intelligence, and the African Union Continental AI Strategy. These initiatives already provide important foundations for ethical, human-centred and development-focused AI governance. However, the added value of the AI Dialogue should be to connect these frameworks more directly to implementation, especially for African and developing countries. In the African context, the Dialogue should support greater alignment between national AI policies across the continent, allowing countries to learn from one another while still responding to their own social, economic, linguistic and infrastructure realities. It should also strengthen partnerships between governments, technology developers, academia, civil society, entrepreneurs and non-technical sectors. The Dialogue could also build upon emerging infrastructure initiatives, such as African AI compute and data centre projects, including partnerships like Cassava Technologies and NVIDIA. These initiatives should not only support technical development, but also create broader opportunities for local businesses, researchers, workers, SMEs and communities to participate in the AI economy. The added value of the AI Dialogue would be to move responsible AI from principles into practical cooperation: funding, infrastructure support, compute access, local language development, skills programmes, governance alignment and inclusive economic participation. For Spiced Paper, the most important value of the Dialogue would be ensuring that AI governance remains human-centred, Africa-relevant and development-focused. It should help countries build AI capability in ways that protect human rights, strengthen local economies, support decent work and ensure that people are not only users of AI systems, but active participants in shaping their future.