While the US, EU, and China argue about how to regulate AI, South Africa just skipped the debate and built a financial safety net for when AI hurts people.

Communications Minister Solly Malatsi on Friday gazetted an 86-page Draft National AI Policy that includes something no other country has tried: an AI Insurance Superfund. Modeled after South Africa’s Road Accident Fund, the state-backed mechanism would compensate individuals and businesses harmed by AI-driven decisions. If an algorithm denies you a loan unfairly, if a deepfake ruins your reputation, if an automated system gets your medical diagnosis wrong, someone pays.

That "someone" would not just be the company that deployed the AI. It would be a national fund designed to ensure victims are not left navigating corporate legal departments alone.

Five New Institutions in One Policy

The Superfund is the headline, but the policy goes further. It proposes creating five entirely new government bodies: a National AI Commission to coordinate policy across government, industry, and civil society. An AI Ethics Board to enforce rules on bias, privacy, and fairness. An AI Regulatory Authority to run audits and issue certifications. An AI Ombudsperson to give individuals a direct path to challenge AI-driven decisions. And the Superfund itself.

Five new institutions. In one policy document. From a country that most AI commentators never mention.

The Radical Admission

What makes this policy genuinely different from every other AI regulatory framework in the world is a single, blunt admission: AI will cause harm, and someone has to pay for it.

That sounds obvious. It is not. The EU AI Act spent years creating risk categories and compliance requirements, but offers no direct financial remedy for individual victims. The White House AI framework published this week is explicitly designed to shield AI companies from liability lawsuits. China’s AI regulations focus on censorship and content control, not victim compensation.

South Africa is the first country to say: we are going to build a financial mechanism that assumes harm will happen and prepays for the consequences.

Will It Work?

That is the hard question. The Road Accident Fund, which this Superfund is modeled on, has been plagued by financial mismanagement and long delays. Creating five new government institutions in a country already struggling with institutional capacity is ambitious to the point of being questionable.

But here is why it matters regardless: the policy establishes a principle. AI companies operating in South Africa would need to contribute to a fund that compensates people their technology harms. That is a fundamentally different starting position from "let the victim sue a trillion-dollar company in court."

The 60-day public consultation period is now open. If this policy survives in anything close to its current form, South Africa will have built the most comprehensive AI accountability framework in the world. Not the EU. Not the US. Not China.

A country with 33% unemployment and rolling electricity blackouts just published a more serious answer to AI harm than any of the countries that actually build AI models. That should make those countries deeply uncomfortable.