Ramaphosa Told BRICS How to Govern AI. He Runs a Country Still Working on the First Industrial Revolution.
The reason a state failing on nineteenth-century logistics and twentieth-century power generation cannot credibly direct the global architecture of the fourth.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
President Cyril Ramaphosa gave South Africa’s country statement at the 18th BRICS Leaders’ Summit in New Delhi recently. He proposed that BRICS establish an international mechanism for the independent scientific evaluation of AI systems, backed by meaningful human control principles, mandatory reporting of serious incidents, progressively stronger safeguards as capability increases, and simultaneous investment in sovereign AI capacity for developing economy countries.
He cited the CEO of Anthropic, Dario Amodei, and Amodei’s September 2026 essay “We Must Pace the Frontier” as evidence that industry itself supports external oversight of AI companies. He compared the required regulatory posture to aviation, pharmaceuticals, nuclear energy and financial services. The speech landed as polished diplomatic performance. It named real risks. It advocated for a defensible international governance architecture. And it was delivered by the president of a country whose rail and port system loses tens of billions of rand a year in trapped commodity exports, whose largest metropolitan municipalities are losing between 45 and 65 per cent of their treated water, whose national AI policy framework was withdrawn in 2025 and remains unpublished, and whose international position on AI turned out to be a subordinate endorsement of two foreign agendas rather than an original South African voice.
CONTEXT AND BACKGROUND
The state of the country the President was speaking on behalf of is worth naming plainly, because it is the country that has to earn any credibility his international position claims. Transnet, the state-owned logistics company that runs South Africa’s rail freight and its ports, has been in operational collapse for the better part of a decade. Freight rail volumes collapsed to 152 million tons in 2023/24, down from 226 million tons in 2017/18.
On the northern corridor alone, which links coal producers to Richards Bay, Transnet requires an estimated R13 billion simply to restore basic functionality, and R65 billion over five years to run the network at design capacity. This is not an emerging risk. It is a first-industrial-revolution technology — steam and rail — that is not currently working. The state has responded by opening the network to private operators, a specific admission that the state’s own capacity is inadequate to run the technology that Britain solved in 1830.
Water infrastructure is worse. EThekwini municipality, which covers Durban, recorded 65 per cent non-revenue water losses for June 2026, with 12,000 reported leaks the municipality could not attend to during a dispute with contracted plumbers. Johannesburg Water reported audited non-revenue water at 44.8 per cent for 2024/25. Nationally, the non-revenue water rate is 47.3 per cent, meaning South Africa loses more than 2.1 billion kilolitres of treated water a year, worth an estimated R26 billion, before it ever reaches a household or business. This is a second-industrial-revolution technology — municipal water reticulation — that was solved in most industrialised cities by the 1920s. It is not currently working in South Africa’s largest metropolitan areas.
On the AI governance dimension of the speech, the position is worse. South Africa withdrew its draft National AI Policy Framework in 2025 and is now targeting January 2027 for a revised version. The country has no operational institution equipped to perform the independent scientific evaluation of AI systems that Ramaphosa proposed BRICS should establish internationally.
There is no dedicated regulator. There is no dedicated research funding programme of scale. There is no legislative framework beyond the general provisions of the Protection of Personal Information Act. There is no timeline for creating any of them beyond the still-abstract 2027 policy framework target.
The comparison the President drew to aviation, pharmaceuticals and nuclear regulation is instructive in the specific way he probably did not intend. The Civil Aviation Authority regulates South African aviation to international ICAO standards, and the country’s air safety record is defensible. SAHPRA regulates pharmaceuticals to substantially international standards. The National Nuclear Regulator oversees Koeberg and has for decades. These are all regulatory institutions that took years to build and that operate on the basis of institutional capabilities the state has invested in. South Africa has nothing equivalent for AI. There is no institution, no mandate, no capability, and no timeline for creating any of them. The president of a country whose regulatory posture on AI is at that stage of development cannot credibly propose to BRICS that the international community should adopt a governance mechanism South Africa has not attempted to build at home.
INSIGHT AND ANALYSIS
The rhetorical move that most exposes the credibility gap is not just that Ramaphosa cited Amodei. It is what happens when the two subordinate endorsements the speech actually made are looked at together. Ramaphosa’s proposal for an international AI evaluation mechanism was tabled inside the broader BRICS agenda on AI cooperation, an agenda being led by Chinese President Xi Jinping. In the same speech, Ramaphosa cited Anthropic CEO Dario Amodei — one of the most senior American AI industry figures — as the credibility anchor for the specific oversight architecture. The speech was, in specific terms, a South African endorsement of an American AI executive’s regulatory vocabulary applied to a Chinese-led BRICS framework. Neither of the two major AI power blocs received a distinctively South African contribution. Neither received anything the country’s own institutional capacity had produced. What both received was subordinate alignment.
That is the sharper diagnosis, and it is more damaging than aspirational overreach. South Africa has substantive domestic capacity on this specific subject. The Centre for AI Research (CAIR), a Department of Science and Innovation-funded national research network established in 2011, has research groups specifically on the ethical and legal aspects of AI, the foundations of machine learning, and AI for development. CAIR operates across five universities and hosts the annual Southern African Conference for AI Research.
The University of Cape Town’s Artificial Intelligence Research Unit hosts CAIR and is positioned by the state as “a key instrument for building South Africa’s foundational capacity for AI research in alignment with the government strategy over the next 10 years.” UNESCO’s readiness assessment names the CSIR, the Centre for the Fourth Industrial Revolution, CAIR, and the South African Research Chairs Initiative as the specific institutional apparatus available for a serious South African AI position. These are institutions the state has funded. They are institutions the President could have cited at BRICS. He did not.
The reveal is not that South African AI thought does not exist. It is that the state has funded it, then chosen to bypass it for a foreign-anchored international position. The president of a country reached past the domestic institutional capacity his own government has paid for, to borrow Silicon Valley’s regulatory vocabulary, in support of a Beijing-led framework. That is not intellectual dependency by accident. It is intellectual dependency by executive choice.
The BRICS speech is the same error at the level of statecraft. The President named the sophisticated future risks of AI. He said nothing about the present risks of a country that cannot move commodities to its ports, cannot deliver water to its metropolitan populations, and cannot publish a policy framework for the technology he was proposing to help govern internationally. Those two categories of risk are not separate. They are the same category, sequenced differently. A state that cannot govern the earlier technologies inside its own borders does not have the credibility to lead the international debate on the later ones, and a state that borrows both its regulatory vocabulary and its diplomatic framework from foreign powers has not developed the intellectual independence to lead any debate at all.
IMPLICATIONS
For South African policymakers, the immediate implication is that the country’s international AI position has to be built on a credible domestic position, and that the timeline for the domestic position has to be accelerated substantially. The January 2027 target for a revised National AI Policy Framework is inadequate given the international ambition the President has now signalled. Bringing that date forward by twelve months, publishing a serious framework by end 2026, and committing to legislated safeguards in 2027 rather than only a policy framework, would be an act of governance capacity the President’s speech implied but did not deliver. Its absence will read, in retrospect, as an admission that the international speech was performance rather than commitment. The framework that eventually appears must draw specifically on CAIR, on the AIRU, on the C4IR, on the National Research Foundation’s SARChI Chairs, and on the civil society legal centres that have been doing this work for years. If the state is willing to fund these institutions, it should be willing to use them.
For South African corporate leaders, the implication is that the international debate on AI oversight will be shaped by BRICS whether or not South Africa is credibly present in it, and that domestic corporate governance on AI cannot wait for a national framework that is at least 15 months away. Boards that have not yet asked what their organisation’s exposure to unregulated foreign AI systems looks like, what their compliance position under Popia is on transborder data flows into those systems, and what their fallback capability is if a critical AI system fails, are now materially behind on their governance obligations. The Information Regulator has been signalling that the direction of enforcement is toward earlier action rather than later. Boards should assume the framework arrives sooner than the President’s speech implied, not because he will make it happen, but because the Information Regulator will act with or without one.
For South African commentators, myself included, the implication is not that the intellectual work has not been done. It has been done. What has not happened is the state engaging with it. The work now is to make that scholarship impossible to bypass, so that the next serving president who addresses BRICS on AI has a body of domestic thought to draw on rather than the temptation to endorse one foreign vocabulary while supporting another foreign framework.
CLOSING TAKEAWAY
Ramaphosa’s BRICS speech was polished diplomatic performance without domestic foundation. It named real risks. It advocated for a governance architecture the country giving the speech cannot yet build at home. It cited an American AI CEO not because there is no South African voice on this subject, but because the state has chosen not to engage the domestic voices it has funded. It endorsed a Chinese-led BRICS framework rather than proposing a distinctively South African one. It was delivered by a president whose country cannot move its own commodities to its own ports, cannot deliver water to its own metropolitan populations, and cannot publish a national policy framework for the technology his speech proposed to help govern internationally. That is not statecraft. It is subordinate alignment performed at an international forum by a country whose domestic record demonstrates it has not yet earned the voice the speech claimed, and whose international position turned out to be the endorsement of two foreign agendas rather than the articulation of a South African one. South Africa’s presence at BRICS is legitimate and its participation in AI governance discussions is strategically necessary. What is not legitimate is the pretence that the country is ready to lead a debate it has not begun to have at home, using a vocabulary it has not developed, inside a framework it has not authored. The work to earn that voice is substantial. It should have started years ago. It can no longer be postponed by another speech.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments