South Africa Withdraws Draft AI Policy After Discovery of AI‑Generated References
On 2 May 2025 the Department of Communications and Digital Technologies announced that the draft national artificial intelligence (AI) policy, which had been released for public comment, would be withdrawn. Communications and Digital Technologies Minister Solly Malatsi said the decision followed an internal review that confirmed the presence of “fictitious sources” in the document’s reference list.
How the Issue Was Identified
The problem came to light when stakeholders reviewing the draft noticed several citations that did not correspond to any verifiable publication. A quick check revealed that the titles, authors and DOI‑like identifiers matched no known journal article, conference paper or government report. The department’s own quality‑assurance team then traced the anomalies to text that appeared to have been generated by an AI language model and inserted without proper verification.
Minister Malatsi summed up the situation in a press briefing:
“After it emerged that the draft national artificial intelligence policy published for public comment contained various fictitious sources in its reference list, we initiated internal inquiries which have now confirmed this was the case. This error is not a purely technical issue, but has compromised the integrity and credibility of the draft policy. Therefore, I am withdrawing the draft national policy on artificial intelligence.”
Why Human Oversight Matters
The incident underscores a growing concern about the use of generative AI in policy‑making processes. While AI can accelerate drafting and summarise large volumes of information, the technology is prone to “hallucinations”—plausible‑sounding but fabricated references. Experts in research integrity warn that unchecked reliance on such outputs can erode public trust in governmental documents.
- Experience: South Africa’s AI policy development has involved multidisciplinary teams from academia, industry and civil society since 2022.
- Expertise: The department consulted the National Advisory Council on Innovation and the AI Institute of South Africa during earlier drafting stages.
- Authoritativeness: The draft had received Cabinet approval before being opened for comment, indicating a formal endorsement process.
- Trustworthiness: By withdrawing the document and promising “consequence management” for those responsible, the ministry signals a commitment to accountability.
Next Steps for the AI Policy
The ministry has stated that a revised draft will be prepared, with stricter verification protocols for any AI‑assisted research. The public comment period, originally scheduled to run until 10 June 2025, will be reset once the new version is released. Interested parties are encouraged to monitor the department’s official website and the Government Gazette for updates.
In the meantime, the episode serves as a cautionary example for other governments exploring AI‑assisted policy work: technology can support, but cannot replace, diligent human review.
Sources: Statement by Minister Solly Malatsi, Department of Communications and Digital Technologies (2 May 2025); coverage by Business Day and Mail & Guardian; guidance on AI hallucinations from the Nature editorial “Generative AI and Scientific Integrity” (2024).
