South Africa’s national AI policy framework has officially hit an emergency operational reset.
In May 2026, communications minister Solly Malatsi appointed an independent expert panel to salvage the country’s withdrawn draft policy after a flawed bibliography exposed fake, machine-generated source citations. We break down the structural timeline behind this revised governance rollout and what these redrafted guardrails mean for your digital business compliance.
South Africa’s AI policy gets a second chance
South Africa wants an AI policy that does more than tick a governance box. It has to help local companies experiment with the technology, while also giving people some protection when AI starts making decisions that affect their work, money, data, or access to public services.
That’s the ambition, at least.
The problem is that the first version didn’t land cleanly. Published in April 2026, the draft tried to put South Africa forward as a serious AI player on the continent. It also raised the big questions we’d expect: who checks AI systems, how jobs may change, what happens to personal data, and how the government keeps the technology from causing real-world harm.
Then the credibility problem arrived.
The department withdrew the earlier draft after reports found fictitious references in the document. Reuters reported that some of those sources seemed to be made by AI, which made the problem even bigger because the policy was supposed to show South Africa how to use and control AI.
That irony is hard to miss.
What happens next?
The government has set up a team of seven experts to check the draft. They will remove any information and suggest what should be kept, changed or removed.
Jeanette Morwane, acting deputy director-general at the Department of Communications and Digital Technologies said the updated policy will be sent to Cabinet by November 2026. People can give their comments on the policy by January 2027.
Here’s the working timeline:
| Stage | Target |
| Expert panel review | 2026 |
| Revised policy to Cabinet | November 2026 |
| Public comment process | January 2027 |
| Final policy work | After public input |
The department has also placed two officials on precautionary suspension while it investigates what went wrong. Director-General Nonkqubela Jordan-Dyani described the incident as “highly regrettable” and said withdrawal was needed to protect the department’s credibility.
Why this matters beyond government paperwork
AI policy can sound dry. It isn’t.
A national AI policy seems like an idea until you think about all the places where AI is already being used. It can affect how companies look at the resumes, how schools use AI tools, how police or private companies use recognition and how government departments use machines to handle public services.
The April draft also tried to sketch out the machinery behind that system. It said there should be groups like a National AI Commission, an AI Ethics Board and an AI Regulatory Authority. The April draft also talked about helping AI development with things like tax breaks, grants and subsidies.
That’s why the rewrite matters outside Pretoria. A clearer policy could affect a fintech startup in Cape Town, a bank in Sandton, a university research lab, or a small business testing AI for customer service. The rules won’t just sit in a PDF. They’ll shape who can use AI, how they use it, and what happens when something goes wrong.
For more context, Memeburn has also covered how South African businesses face growing legal risks as AI adoption soars, especially around bias, contracts, liability, and compliance.
This also matters because South Africa can’t afford an AI policy that only works for big companies with legal teams and deep technical skills. Small businesses, schools, startups, and public offices will need rules they can actually understand and follow.
If the policy becomes too vague, companies may move fast without enough safeguards. If it becomes too heavy, smaller players may avoid AI altogether.
That balance will be tricky. South Africa needs space for experimentation, but it also needs clear guardrails around privacy, bias, accountability, and public trust. The January 2027 draft will need to show that the government has learned from the first mistake — not just by fixing citations, but by building a policy process people can believe in.
The fake citation problem is bigger than one document
This isn’t just about a bad bibliography.
Generative AI tools can produce convincing-looking text that turns out to be false. When that happens in a school essay, it’s a problem. When it happens in a national policy document, it becomes a trust issue.
The department’s mistake shows why human oversight matters. AI can speed up research, summarise documents, and help draft policy language. But humans still need to verify every source, every claim, and every legal reference before publication.
That matters even more in government, where policy documents help guide future laws, budgets, institutions, and enforcement.
South Africa still needs an AI plan
The withdrawal doesn’t mean South Africa can ignore AI policy until 2027.
Companies are already using AI in areas such as customer service, hiring, insurance, banking, media, cybersecurity and software development. Public institutions are also looking into AI for delivering services, research, forecasting and administration.
Without clear rules, South Africa faces two problems.
First, weak oversight could put people at risk of being exposed to systems, privacy failures and automated decisions they don’t understand. Second, excessive delays could slow down local innovation. While other countries are moving faster with AI frameworks, South Africa might fall behind if it does not act quickly.
The earlier policy framework said South Africa wanted AI to support economic growth, public benefit, and responsible innovation. The new draft now needs to prove that ambition can survive basic scrutiny.
The trust test starts now
The government’s biggest challenge isn’t just rewriting the draft. It needs to rebuild trust.
That means the next version must show where its claims come from, how experts shaped the document, and how ordinary South Africans can respond during public consultation.
The new panel gives the process a cleaner starting point. But the January 2027 draft will face tougher questions because the first version failed in such a public way.
South Africa wants to lead on AI in Africa. That’s still possible.
But the next draft can’t just sound smart. It has to be checked, sourced, and strong enough to govern the technology it describes.
FAQs
When will South Africa release the revised AI policy?
South Africa is now going to open the Artificial Intelligence policy, for public comment in January 2027. Before that happens, the Department of Communications and Digital Technologies wants to send the updated draft to the Cabinet by November 2026. So, the next big moment to watch is not the final law yet — it’s the public consultation draft.
Why did South Africa withdraw the earlier AI policy draft?
The earlier draft ran into trouble because it included fake and possibly AI-generated references. That’s a serious problem for any government document, but it looks even worse when the document is about artificial intelligence. The withdrawal gives officials a chance to rebuild the policy with proper sourcing, stronger checks, and more public trust.
What could the AI policy change for South Africans?
It can make rules for AI in business, government, education, jobs and data privacy. It can also be guided by this when it comes to how South Africa helps AI startups and keeps citizens safe from bad automated systems.
