Minister Solly Malatsi says Cabinet will have a cleaner document in hand by the end of the financial year. The date has slipped, and the country remains without a national AI policy.
JOHANNESBURG, October 26, 2026. South Africa’s government says it will have a replacement national artificial intelligence policy before Cabinet by the end of March 2027, nearly a year after the first attempt was approved with great fanfare and then pulled within weeks because its reference list contained sources that did not exist.
Communications and Digital Technologies Minister Solly Malatsi has said he is confident Cabinet will finalise the new policy before the end of March, with public consultation expected to begin in April, according to TechCabal. The commitment follows one of the more embarrassing episodes in African AI governance, and the new deadline is later than the one his own department gave Parliament earlier this year.
How the first draft fell apart
The Draft National Artificial Intelligence Policy was approved by Cabinet on March 25, 2026, with a further special sitting on April 1, before being released for public comment. It was published in the Government Gazette on April 10, opening a 60 day consultation period with written input invited until June 10.
The document was meant to be a landmark. It was built around six strategic pillars, including capacity and talent development, responsible governance and human centred deployment, and it proposed a phased three year implementation plan running through the 2027/28 financial year. Its stated vision was AI for inclusive economic growth, job creation and a developing Africa.
That ambition collapsed over a single weekend. A News24 report revealed that the draft’s reference list included fabricated research. Within days, an internal check confirmed the problem, and on Sunday, April 26, Malatsi announced that he was withdrawing the policy. Coverage of the announcement appeared the following day.
The minister did not soften his language. He said the failure was not a technical glitch but a blow to the integrity and credibility of the document, and that his department had fallen short of the standard expected of the institution leading the country’s digital policy. He called the most plausible explanation that AI generated citations had been included without proper verification, and promised consequence management for those responsible for drafting and quality assurance.
The fallout
The consequences have been both internal and international. TechCabal reports that two officials were suspended over the affair, that one of them has since resigned, and that disciplinary proceedings are continuing against the other.
Abroad, iAfrica reported that the World Bank’s World Development Report 2026 left South Africa off its map of economies with a national AI strategy, citing the April withdrawal directly, even though South African institutions, including Malatsi’s own department, are credited as contributors to the report. The Bank’s report has not been reviewed independently for this article, so the claim rests on that reporting.
At home, the department was exposed to a harder question: how a document that demanded human oversight of AI had itself escaped human scrutiny. Malatsi has been candid about where his office’s checks stopped. Speaking at GovTech 2026 in Durban, he told Daily Maverick that his team had pushed hard on the substance of the drafts but “didn’t interrogate the bibliography,” because it assumed anyone doing the work honestly would not rely on AI without saying so.
From that experience he has drawn a standard. Using AI in research and drafting, he argues, becomes unethical only when the use goes undeclared.
Rebuilding with a panel
The repair work began in May. On May 14, government appointed an independent AI review panel to rebuild the withdrawn document and recommend revisions before it returns to Cabinet. iAfrica reports that the panel is chaired by Professor Benjamin Rosman of the Wits Mind Institute, and Malatsi has publicly praised its calibre.
Malatsi has also insisted that the withdrawn draft was not worthless. He has said it contained credible provisions, including protections for South Africans’ rights and dignity online, and the replacement is expected to carry that thinking forward while addressing the weaknesses the scandal exposed.
On the philosophy of regulation, the minister has warned against the reflex to regulate everything at once. In his Durban remarks, he argued that rules should be used to enable sectors to grow while still safeguarding the public, not to choke off opportunity.
A moving deadline
The March 2027 target is the latest in a series of shifting dates. In May, TechCabal reported that the policy had been pushed back, with a revised framework expected to open for public comment in January 2027. Acting deputy director general Jeanette Morwane had also put the redraft before Cabinet in November 2026, with January as a fallback for consultation.
Malatsi’s September timeline is later. iAfrica noted that a shift from November 2026 to March 2027 amounts to at least four months, and that the department had not formally announced the change.
The wording also matters. Malatsi described the goal in September as having a draft policy ready to present to Cabinet by the end of the financial year, one that could be free of what he called avoidable stains. With public consultation not expected until April, the March milestone appears to be a Cabinet cleared draft heading out for comment, rather than a finished policy ready to take effect. Readers should not assume the country will have a final, adopted framework in the first quarter of 2027.
Why the timing matters
South Africa is not short of AI activity. Businesses and public institutions are adopting AI powered systems at pace, and the absence of a formal national policy leaves them without clear guidance from the state.
The pressure to act is growing across the continent as well. In Kenya this month, a Senate committee recommended that lawmakers not proceed with a proposed AI Bill, arguing that a national policy should come first. South Africa has the opposite problem: it knows the policy must come first, but it has yet to deliver one it can stand behind.
For now, the country’s credibility on AI governance rests on a date and a promise. Whether March 2027 holds, and whether the document that reaches Cabinet earns the trust the first draft lost, will say a good deal about how seriously a continental leader in technology can be taken on the rules for the technology itself.
Sources: SAnews (April 2026), Mail & Guardian (April 27, 2026), ITWeb (April 2026), TechCabal (May 26 and October 15, 2026), iAfrica.com (April and September 25, 2026, reporting Daily Maverick), Innovation Village (October 21, 2026).
