As the AI race accelerates the conversation tends to concentrate on infrastructure, compute and data. These matter, but they are not the whole contest. National AI strategies are multiplying worldwide, and the rules are becoming part of the race itself. Earlier this year, at the Global AI Impact Summit in New Delhi, 92 countries and organisations endorsed the New Delhi Declaration, the broadest consensus on AI to date. Two things from Delhi should stay with us. First, India, a Global South country, did not wait for permission. It convened the world and set the terms. Second, like most declarations, New Delhi’s is voluntary. It becomes real only when countries build the capability to implement it. Consensus sets direction; institutions must deliver, and at speed. That is the race South Africa has entered: an implementation race, won or lost not in legislation but in institutions, professions and partnerships. AI governance does not become real when a principle is agreed or a policy is published. It becomes real when it changes a decision, a workflow, a control or an outcome, and when an organisation can produce evidence of that change. The global conversation is moving from what the principles should be to who can operationalise them. Countries and companies are being sorted into those that can implement and those that can only aspire. Regulatory certainty has become a competitive asset: capital, talent and compute are flowing to jurisdictions where the rules are clear, workable and predictably enforced. That is the prize now in play for South Africa. This country deserves real credit for the moment it has created. By placing its AI policy framework under an independent expert panel before drafting, it has chosen rigour over haste. The EU, for example, legislated first and learnt afterwards (to its cost); South Africa can learn first and build a framework that is stronger, more workable and more investable from day one. But we need to use this advantage with a sense of urgency, because the window will not stay open for long — AI is developing at a rate rarely (never?) before seen in technological history. I offer a view from inside a global technology group. Naspers and Prosus operate in more than 70 markets, and our teams live with regulatory divergence every day: similar products, similar technologies, but a patchwork of jurisdictions that have each made different choices about what is permitted, what must be disclosed and who carries liability. It has taught us one lesson above all: governance regimes work best when implementers are in the room. When regulators understand how organisations build, buy and deploy these systems and how they enable consumers, we can together shape appropriate rules. Without that understanding and joint enterprise, we risk rules that are disconnected and ultimately unworkable. In addition, workable national frameworks, necessary as they are, are only the beginning. Whether AI delivers its full potential for this country (a potential that we can’t yet begin to understand) will not be decided in the Government Gazette. It will be decided in the implementation layer: regulators who engage with institutions as they genuinely operate; company secretaries and governance professionals who turn obligations into controls and accountability that survive contact with a real organisation; and legal and policy teams that design the workflows through which AI is deployed. All done in a way that ensures that AI’s potential is not limited through irrelevant or inappropriate constraints. It has taught us one lesson above all: governance regimes work best when implementers are in the room. When regulators understand how organisations build, buy and deploy these systems and how they enable consumers, we can together shape appropriate rules. Let me highlight my own profession. The caricature says lawyers treat AI as another risk to be slowed down. That is nonsense. Law is now at the forefront of the professional services sectors that are genuinely embracing AI inside core workflows: starting with research, drafting, due diligence and contract review, the profession is now using AI to transform how law is practised. When the profession shaped to manage risk adopts a technology at this pace, it signals to every other function that responsible adoption is possible. Legal teams that engage early earn the seat where AI decisions are made and shape those decisions before they harden. The pattern is clear: the nations that get AI governance right will not be the ones with the most elegant legislation. They will be the ones that treat governance as a capability, founded on deep understanding and built in institutions, professions and people. The world watched this country negotiate a transition to democracy that much of the world believed impossible and a constitution the world still studies, and forge institutions out of more than conviction and goodwill. So, when I say that South Africa can get AI governance right I am not offering polite encouragement. Building under pressure and against the odds is our inheritance. Let us be clear-eyed about the stakes: for us as a country, AI policy is not a compliance exercise, it is a critical element of an inclusive economic strategy that has AI at its core. Done well, it becomes a magnet for investment, a platform for our entrepreneurs, a multiplier for our industries and a bridge over the inequalities that have held us back. Getting it wrong will set us on a path that is too horrible to contemplate. The world is yet again deciding how a transformative force will be governed, and the rules are still being shaped. South Africa should not be content with a seat on the sidelines. We have the credibility, the institutions and the talent to help lead the Global South in shaping these rules. • Tudor is group general counsel of Naspers and Prosus and a member of the group management team.