Communications Minister Solly Malatsi has ordered the withdrawal of South Africa’s artificial intelligence policy draft after fabricated, AI-generated references were found embedded in the official document. The minister has commissioned an independent expert panel to develop a replacement framework, with completion targeted for 2027.
The withdrawal is a direct accountability failure at the ministerial level. It exposes gaps in the verification and quality-control processes that govern how technical policy documents are reviewed before publication, raising pointed questions about institutional oversight.
The decision to bring in an independent expert panel represents the government’s attempt to restore rigor to a process that clearly broke down. By externalizing the rewriting effort, the department is acknowledging that internal review mechanisms were insufficient to catch fictitious references in a document intended to guide national policy. The 2027 completion date signals a deliberate, thoroughness-first approach rather than a rush to repair reputational damage quickly.
Meanwhile, the policy failure lands at a particularly costly moment. Across Africa, governments are actively competing to attract investment in artificial intelligence, digital infrastructure, and technological innovation. South Africa had positioned itself as a prospective continental leader in this space. The discovery of unverified AI-generated content in an official policy document complicates that claim and hands competitors an opening.
Technology experts have flagged the broader institutional implications. When AI-generated content passes through official channels without adequate human validation, the integrity of the policy record is compromised. For a government seeking to project sophistication in the AI sector, that kind of error carries weight that extends beyond a single document.
The incident has also sharpened debate about the safeguards institutions need when deploying AI tools internally. The question is not simply whether AI was used in drafting, but whether the oversight structures existed to catch what the technology got wrong. In this case, they did not.
South Africa’s formal AI policy framework will now remain in development for several additional years. During that period, the country will continue competing for investment and talent without a finalized national policy to anchor that effort. Whether the independent expert panel can deliver a document that restores confidence in the government’s technical competence, and whether the institutional controls that failed once have genuinely been reformed, remains the open question the 2027 deadline will eventually have to answer.