South Africa Hands AI Policy Over to Expert Panel After Draft Withdrawal

South Africa has withdrawn its draft national artificial intelligence policy after several references in the document were found to be unverifiable and linked to AI-generated hallucinations, triggering concerns over the credibility of the country’s AI governance process.

The controversy emerged after media reports revealed that some citations included in the policy document could not be traced to legitimate sources.

Speaking before Parliament’s Portfolio Committee on Tuesday, Solly Malatsi acknowledged that the government only became fully aware of the issue after the media investigation exposed the errors.

According to Malatsi, at least six references contained in the draft policy were attributed to fabricated or AI-generated information, prompting the department to immediately launch an internal review.

The minister admitted that the incident had damaged public trust in both the policy document and the broader government policymaking process.

He described the situation as a major oversight, particularly regarding the undisclosed use of artificial intelligence during the drafting and referencing stages of the policy.

Malatsi said the withdrawal of the draft was necessary to protect the integrity of the process and restore confidence in South Africa’s approach to AI regulation and governance.

The government has since placed two departmental officials on precautionary suspension pending the outcome of a formal investigation into how the fabricated references were included in a Cabinet-approved document without being detected earlier.

According to the minister, the issue raised serious concerns not only about inaccurate citations but also about the credibility of the entire policy framework.

As part of efforts to rebuild trust and strengthen oversight, the Department of Communications and Digital Technologies has established an independent AI advisory panel made up of experts in artificial intelligence, governance, academia, and law.

The panel will be responsible for reviewing the policy document, verifying references, and guiding the next phase of the drafting process.

The advisory body is chaired by Benjamin Rosman of the MIND Institute at Wits University and includes experts such as Vukosi Marivate, Alison Gillwald, Jabu Mtsweni, Tshepo Feela, Advocate Lufuno Tshikalange, and attorney Heather Irvine.

Malatsi stated that the panel is expected to finalise its terms of reference in the coming weeks and develop a revised roadmap for the policy process.

The South African government now aims to complete and deliver a revised national AI policy before the end of the current financial year, although officials noted that Cabinet procedures and public consultation timelines could influence the final schedule.

The incident has sparked wider conversations about the growing use of artificial intelligence in policymaking, research, and governance, particularly the risks associated with relying on unverified AI-generated content in official state documents.

TNAM
Edited By Egwu Patience Nnennaya