South Africa just published an 86-page draft AI policy that proposes creating seven new government bodies to oversee artificial intelligence. A National AI Commission. An AI Ethics Board. An AI Regulatory Authority. An AI Ombudsperson Office. A National AI Safety Institute. An Integrated AI-Powered Monitoring Centre. And a dedicated funding mechanism to pay for all of it.

Communications Minister Solly Malatsi gazetted the policy on April 10, opening a 60-day public comment window. The stated goal: position South Africa as a continental leader in AI innovation while addressing ethical, social, and economic challenges. The unstated reality: this might be the most ambitious AI governance proposal any country has attempted. Whether that is brave or delusional depends on who you ask.

The Scope Is Staggering

The draft outlines a three-phase implementation plan. Phase one runs through September 2026 and focuses on building the institutional framework. Phase two extends through March 2027 and covers publishing AI guidelines, implementing high-risk use case regulations, and drafting requirements for medium and low-risk applications. Phase three is full enforcement.

Reuters reports the policy builds on six pillars: governance, innovation, skills development, infrastructure, ethics, and international cooperation. It also includes the AI Superfund concept that South Africa previewed earlier, creating a financial mechanism to compensate people harmed by AI systems.

Critics Are Already Circling

TechCentral called it "a bureaucrat's dream." And honestly, they are not wrong. South Africa struggles to keep its existing regulatory agencies funded and functional. Adding seven new ones for a technology that barely any South African company is building feels like creating a fire department for a town that does not have buildings yet.

The counterargument is more interesting. The EU spent years building the AI Act and companies are now racing to deploy before enforcement kicks in. Europe's rules are already being gutted under industry pressure. South Africa watched that happen and decided to try building enforcement mechanisms before the technology arrives at scale, not after.

The Real Question

Compare this to America, where Congress has introduced 47 AI bills and cannot agree on a definition of AI. Or the EU, which built a comprehensive framework and is now weakening it under lobbying pressure. South Africa's approach is the opposite extreme: regulate everything, create oversight for everything, fund institutions for everything.

Neither extreme works. But at least South Africa is trying something while the world's largest economies are still arguing about whether AI needs rules at all. The Citizen noted that the policy signals "the era of move fast and break things is over in South Africa." That is true. The question is whether seven new agencies can move fast enough to matter.

Public comments are open until June 10.

First reported by Reuters. Additional reporting from TechCentral, The Citizen, and SABC News.