South Africa’s draft artificial intelligence policy has been withdrawn after government officials discovered it contained fabricated references and content generated by AI itself. Communications Minister Solly Malatsi has announced that an independent expert panel will now lead a complete rewrite, with a revised framework targeted for delivery by 2027.
The policy failure is immediate and concrete. A document meant to guide national AI strategy was found to contain fictitious citations, raising direct questions about the verification procedures, or absence of them, that allowed the draft to advance. The government’s response has been to scrap the document entirely and commission a fresh process under independent oversight.
That 2027 deadline is the operational problem. South Africa will spend an extended period without a formal, vetted AI policy while the rewrite proceeds. Meanwhile, neighboring countries and international investors are actively evaluating where to direct resources for AI research, talent development, and infrastructure projects. The gap between now and 2027 is not neutral time. It is time during which competitors move forward.
The timing compounds the credibility damage. African nations are competing intensely to position themselves as hubs for AI investment and digital infrastructure development. A policy document marred by fabricated references sends a direct signal about institutional capacity and governance standards to the technology companies, research institutions, and venture capital investors that South Africa is trying to attract.
The incident has also sharpened a broader conversation about the risks of using generative AI tools in high-stakes policy work. Technology professionals have emphasized that proper verification protocols, editorial oversight, and domain-specific expertise must remain central to any policy development process, regardless of the efficiency gains AI might promise. The South Africa case makes the argument in the starkest possible terms: AI tools cannot substitute for human accountability when the output carries national strategic weight.
The decision to withdraw the flawed document and bring in an independent panel reflects an acknowledgment of how serious the problem is. Whether that acknowledgment translates into a technically sound, credible framework by 2027 is the question the panel now has to answer.
For South Africa, the reputational damage in the continental AI race has already been inflicted. The more pressing uncertainty is whether the independent process can produce a policy rigorous enough to rebuild confidence among domestic stakeholders and international observers, and whether that confidence can be restored quickly enough to matter in a field where positioning decisions are being made now.