South Africa’s draft national AI policy is back on the drawing board after Communications Minister Solly Malatsi moved to repair the damage caused by a drafting scandal that embarrassed government and raised serious questions about how policy is being made in the age of generative AI. Malatsi told Parliament on Tuesday that an independent panel of researchers, lawyers and governance specialists has been appointed to rebuild the document, after the first version was pulled when fabricated references were discovered in its bibliography.
The minister used his budget speech for the Department of Communications and Digital Technologies’ R2.55-billion allocation for 2026/2027 to address the issue head-on. He said the department could not talk about policy reform without confronting what happened to the AI draft, which was prepared with the irresponsible use of generative AI. In his words, South Africa deserves better.
Malatsi said the department will now enforce an internal responsible AI use policy and overhaul the way policy documents are developed so that a similar failure does not happen again. The intervention comes as the government tries to rebuild credibility on a file that should have positioned South Africa as a serious player in the global AI conversation, but instead became a cautionary tale about cutting corners with the very technology it seeks to regulate.
The new panel will be led by Wits University AI researcher Benjamin Rosman, who was recently named one of Time’s 100 most influential thinkers in AI in 2025. That appointment alone signals the seriousness of the clean-up operation. Our view is that government needed more than a cosmetic fix here; it needed a team with enough academic and practical weight to reassure the public that the redraft will be grounded in evidence, not convenience.
Alongside Rosman are Professor Vukosi Marivate of the University of Pretoria, Alison Gillwald of Research ICT Africa, attorney Heather Irvine, Tshepo Feela, the CSIR’s Jabu Mtsweni and Lufuno Tshikalange. The mix of expertise is notable: AI research, law, policy and digital governance all sit around the same table, which is exactly what a national AI framework should have had from the start.
Malatsi told MPs that the panel’s job is to ensure the revised policy reflects the best available evidence and is aligned with South Africa’s priorities before it goes back out for public comment. No deadline has been announced for the new draft, which means the country remains in a holding pattern on a policy area that is becoming increasingly urgent.
The original draft had been approved by Cabinet on 25 March and gazetted for public comment on 10 April, only to be withdrawn within roughly two weeks. That fast reversal reflected the scale of the embarrassment, especially once the fake citations began to come under scrutiny. What should have been a routine public policy consultation quickly turned into a national talking point about AI literacy in government.
Draft national AI policy dragged into crisis after fake citations exposed
The core of the scandal was not just that mistakes were found, but that the document contained what appeared to be fabricated academic references. At least six of the 67 entries in the reference list were traced to journals that either did not exist or had never published the works cited. Editors of the South African Journal of Philosophy, AI & Society and the Journal of Ethics and Social Philosophy separately confirmed to News24 — which first reported the story — that articles attributed to their publications had never appeared.
That revelation landed hard. It was not simply a formatting problem or a missing source. It suggested the policy was built on material that could not be verified, undermining confidence in the document’s integrity. Malatsi acknowledged as much in Parliament, describing the use of generative AI in the drafting process as irresponsible and saying withdrawal was the only viable option.
The political fallout has been just as sharp. ANC MP Khusela Diko, who chairs Parliament’s communications portfolio committee, called for the policy to be scrapped entirely and accused Malatsi of looking for a “scape-bot” instead of taking responsibility. On the other side of the aisle, Public Works Minister Dean Macpherson, a fellow DA cabinet minister, publicly defended Malatsi.
Outside government, the debate has also exposed deep divisions about how South Africa should approach AI regulation. Technology investor Stafford Masie earlier warned in an open letter that the draft risked “regulating away” the country’s role in the global AI economy by placing too much emphasis on governance and not enough on infrastructure investment and innovation enablement. That argument still resonates with many in the tech sector, who say South Africa needs a policy environment that encourages development rather than choking it at birth.
The withdrawn draft itself was ambitious, and that ambition may have been part of the problem. It proposed creating seven new institutions, including a National AI Commission, an AI Ethics Board, an AI Insurance Superfund and a National AI Safety Institute. On paper, those structures suggested a serious effort to get ahead of the risks posed by rapidly developing technology. In practice, critics said the framework was too broad, too bureaucratic and too far removed from the real constraints facing the country.
There is also a broader governance lesson here. If government cannot produce a credible AI policy without relying on questionable machine-generated references, it raises uncomfortable questions about capacity, oversight and editorial discipline inside the state. The episode has become a live example of the risks of using AI without proper human verification, especially in a setting where policy can shape markets, investment and digital rights for years to come.
Malatsi also used his appearance before MPs to update them on two other major policy tracks. He said the department is finalising the audio-visual services and media policy, while the Electronic Communications Amendment Bill will be taken forward to modernise licensing and deal with convergence in the sector. Those measures matter, but they are now unfolding in the shadow of the AI debacle.
For now, the absence of a formal national AI framework leaves South Africa in a vulnerable position. AI is already embedded in public administration, corporate workflows and consumer services, yet the country still lacks a clear rulebook to govern how it should be used, tested and deployed. That gap matters, particularly as other markets move faster to set standards around transparency, safety and accountability.
Malatsi insists the new process under Benjamin Rosman and the rest of the panel will put South Africa back on course. The real test, however, will not be the minister’s promise in Parliament. It will be whether the next draft is clean, credible and useful enough to survive public scrutiny without becoming another headline about what went wrong.