South Africa has appointed an independent expert panel to rebuild its draft national AI policy after discovering fabricated academic citations [1].
The move follows a public failure in the government's attempt to regulate artificial intelligence, highlighting the irony of using AI-generated hallucinations to draft a policy meant to govern the technology. This setback forces the government to restart a critical legislative process to ensure the framework is grounded in verifiable evidence.
Communications Minister Solly Malatsi said the panel was appointed on May 26, 2026 [4]. The group consists of AI researchers, lawyers, and governance specialists tasked with reviewing and rewriting the policy [1].
The original draft was withdrawn after it was revealed that some of the academic citations were fictitious [2]. Specifically, six out of 67 citations in the document were found to be fabricated [5].
Malatsi said the inclusion of these hallucinations was a "massive oversight" [6]. The errors prompted an immediate withdrawal of the document to prevent the adoption of a policy based on non-existent research [2].
The government has set a strict timeline for the recovery process. The revised AI policy is targeted for presentation to the Cabinet in November 2026 [3]. Following that review, the government expects to open the policy for public comment in January 2027 [7].
The independent panel will work to ensure the new draft meets international standards of accuracy and governance. This process aims to restore trust in the administration's ability to oversee emerging technologies without falling victim to the very errors the policy seeks to mitigate.
“"massive oversight"”
This incident underscores the 'hallucination' risks inherent in generative AI, where models produce confident but false information. For a government, relying on such tools for policy-making without rigorous human verification can lead to a loss of institutional credibility. The appointment of a multidisciplinary panel of lawyers and researchers suggests South Africa is shifting toward a 'human-in-the-loop' verification model to ensure its legal frameworks are evidence-based.
