South Africa will push for the 81st UN General Assembly to produce a statement committing member states to develop a global framework for governing artificial intelligence, according to its Permanent Representative to the UN, Ambassador Mxolisi Sizo Nkosi.
“It’s also important that a statement should come out of this General Assembly committing the international community to the development of a framework for the global governance of artificial intelligence,” Nkosi told SAnews.
One position, two forums
The request follows President Cyril Ramaphosa’s proposal at the BRICS summit in New Delhi last week, where he called for an international mechanism for independent scientific evaluation of AI — backed by human-control principles, mandatory incident reporting, progressively stronger safeguards as capability increases, and investment in sovereign AI capacity for developing economies.
Taken together, the two constitute a deliberate strategy: pursue the same governance argument through BRICS and the UN at once.
Nkosi framed the risk in terms similar to Ramaphosa’s. “How AI can really have a devastating impact if we do not have the guardrails, we do not have a framework that really guides the operation, the development and the spread of AI,” he said. “This is a matter that has been raised not only by governments but also by tech companies themselves, who have really called for the slowing down of AI development due to what they see as the emergence of uncontrolled models and some rogue behaviour by AI systems.”
Ramaphosa named Anthropic chief executive Dario Amodei in making that argument at BRICS, citing his call for external audits and government oversight of AI companies.
What already exists
South Africa’s request enters architecture the UN has already begun building. The General Assembly established an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance in 2025. South African computer scientist Vukosi Marivate, co-founder of Lelapa AI, sits on the scientific panel. Benin, Burkina Faso, Côte d’Ivoire, Guinea, Mali and Senegal submitted a shared governance framework to the first Global Dialogue in July.
The open question is whether South Africa is seeking to strengthen that existing machinery or to move toward something binding. Nkosi’s phrasing — a framework to guide “the operation, the development and the spread of AI” — suggests the latter, but no text has been tabled.
A consistent doctrine
Across three recent positions, South Africa’s view on AI governance has been coherent.
At BRICS, Ramaphosa argued for a multilateral evaluation mechanism because no single developing country can independently assess frontier systems. At the UN, Nkosi is seeking global rules. And domestically, the Financial Sector Conduct Authority said last week it will wait for the Financial Stability Board’s international principles before writing its own AI rules, because it wants to be aligned.
The underlying premise is the same each time: AI governance only works at a scale larger than one state. That is a defensible position, and one many smaller economies share. It also sits alongside the fact that South Africa’s own national AI policy, withdrawn in April, will not reach public consultation until January 2027.
Development and inequality
Nkosi placed AI alongside South Africa’s other UNGA81 priorities: international peace and security, climate change, UN reform, inequality and the selection of a new Secretary-General.
“Other pillars of our foreign policy are around the promotion of respect for human rights, the narrowing of the gap between the developed and developing countries,” he said.
That framing is where South Africa’s contribution is most distinctive. Western AI safety advocacy has centred on catastrophic risk and model evaluation. Ramaphosa’s BRICS proposal and Nkosi’s UN push pair governance with capacity — arguing that developing countries cannot be asked to accept rules for technology they have no means to build or assess.
UN Secretary-General António Guterres has separately called for stronger international cooperation and common safeguards on AI.
