South Africa Withdraws Draft AI Policy After Fake Citations

According to Reuters and local reporting, South Africa withdrew its draft national AI policy on April 26 after investigators found fabricated academic citations in the document. Reuters and India Today report that at least six of the policy's 67 references were fictitious and that Communications Minister Solly Malatsi described the most plausible explanation as "AI-generated citations included without proper verification." The draft, published for comment on April 10, had proposed institutions such as a National AI Commission, an AI Ethics Board and an AI Regulatory Authority, along with tax incentives and grants, according to Reuters. Industry commentators and The Conversation framed the episode as a reminder about the limits of current generative-AI outputs and the need for stronger human verification.
What happened
According to Reuters and multiple media outlets, South Africa's Department of Communications and Digital Technologies withdrew its draft national AI policy on April 26 after independent checks identified fictitious sources in the document's reference list. India Today and The Independent report that the draft, published on April 10 for public comment, contained 67 cited references of which at least six were determined to be fabricated. Communications Minister Solly Malatsi posted on X that "the most plausible explanation is that AI-generated citations were included without proper verification" and said the lapse "has compromised the integrity and credibility of the draft policy," per Reuters.
Technical details
Multiple outlets, including India Today and Mashable, report that the fabricated entries resembled classic generative-AI "hallucinations" and that commonly available models such as ChatGPT or Google Gemini were suggested as possible sources of the invented citations. Reporting by News24 and regional outlets prompted the recheck of the references, and editors at named journals, including the *South African Journal of Philosophy* and *AI & Society*, told reporters that the cited articles did not exist, per India Today and The Independent.
What the draft proposed
Reporting by Reuters and other outlets shows the withdrawn draft outlined a national framework that included:
- •establishment of a National AI Commission
- •creation of an AI Ethics Board and an AI Regulatory Authority
- •incentives such as tax breaks, grants, and subsidies to encourage private-sector collaboration
These elements were reported as part of the draft but the publications note the policy was pulled before finalization.
Editorial analysis - technical context
Generative models produce plausible-sounding citations and references without grounding in verifiable sources; industry analysis, including commentary in The Conversation, highlights that such hallucinations are a well-documented failure mode. Observed patterns in similar incidents show that automated drafting workflows that do not include robust source-verification steps can allow fabricated content to propagate into formal documents. For practitioners, this episode underscores the operational need for metadata provenance, citation validation, and human-in-the-loop checks when generative outputs are used in policy writing or other high-stakes texts.
Context and significance
Industry context
Governments worldwide are moving to regulate AI while also experimenting with AI-assisted drafting. Reporting frames South Africa's withdrawal as a reputational setback at a moment when the country had sought to position itself on AI governance. The episode feeds into broader public and regulatory debates about the reliability of generative systems and the importance of verification in public-sector outputs. Several outlets, including The Register and Mashable, also connected the incident to a larger pattern of AI hallucinations affecting legal and administrative documents globally.
What to watch
- •Whether the Department of Communications and Digital Technologies publishes a corrected draft with documented verification processes, as Reuters notes officials said consequence management would follow.
- •If parliamentary oversight or audits (reported in local coverage such as BusinessTech) lead to procedural changes for government use of generative tools.
- •Signals from academic journals or professional bodies about how they will respond to being mis-cited, and whether tooling for automated citation validation gains traction among governments and policy teams.
Limitations
All factual claims about the withdrawn draft, the number of fabricated references, and Minister Malatsi's remarks are drawn from the Reuters report and corroborating domestic and international coverage (India Today, The Independent, Mashable, The Register). No source provides an internal forensic report publicly attributing the full drafting process to any specific model or workflow.
Concluding note for practitioners
For practitioners building pipelines that might surface outputs into public-facing documents, this incident reinforces the operational requirement for provenance, reference-checking, and explicit human sign-off before publication.
Scoring Rationale
A national AI policy withdrawal directly affects how practitioners and regulators approach oversight of generative systems. The story is notable for governance practice lessons and has broader implications for verification workflows used in policy and high-stakes documents.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

