Solly Malatsi needs to draft an AI policy for South Africa that shows proof of a human touch.
In March, when Malatsi’s department of communications & digital technologies published the draft national AI policy for public comment, this statement would have seemed innocuous.
The saga is embarrassing for the government and Malatsi’s party, the DA. It will surely feed business school management case studies for years to come and is a good example of journalism in action, with News24 doing the work that broke the story.
But the crisis of credibility now hanging over the communications department highlights two facts that government, business and consumers will have to contend with.
The first is that the erosion of trust caused by AI systems making something up or getting it wrong impacts the user more than the platform. A week after Malatsi withdrew the policy document, hardly anyone is concerned about whether the fictitious research came from ChatGPT, Gemini, Claude, OpenClaw, DeepSeek or Perplexity.
Responsibility, blame and the ensuing embarrassment have fallen on the minister’s shoulders.
As Darren Olivier, partner at law firm Adams & Adams, puts it: “AI can damage it [trust] quickly, and once trust is shaken, recovery can take considerable time. In some professions or industries, it may never fully happen.
AI risk ‘not theoretical’
“As this episode has illustrated, the risk is not theoretical. It is already playing out in practice, and damage to brands, be those personal in the case of a lawyer or their firm or for an organisation or government (as it is in this case), and the damage is tangible.”
The negatives of this situation are quite apparent. Nevertheless, Malatsi should be credited for taking swift, decisive action to withdraw the policy document. While not ideal, he at least demonstrated respect for the country’s policy-making process. Others in his position, whether in the public or private sector, may have hesitated or tried to delay the response.
Ultimately, Malatsi’s dilemma shines a light on the second, now inconvenient, truth about AI. The technology is smart, some say intelligent, but it needs human oversight.
Fabricio Bloisi, the man leading South Africa’s largest technology company, Naspers/Prosus, has — over the last year — noted that the world of work is changing to the point where managers are leading or managing hybrid teams. That is human employees, plus AI assistants and agents.
The inconvenience of this truth is that AI is not the “set it and forget it” technology that many thought it would be, at least not in its current form. Its use carries the added burden of having to ensure that the output is valid. While it can and has taken large amounts of tedious “graft” from humans, it has added responsibility to those making use of it, particularly organisation leaders.
Since ChatGPT ushered in mainstream adoption of AI in November 2022, experts have advocated for this “human in the loop” approach.
Until AI gets to a point where society can trust its output in the same way that answers on a simple calculator are not questioned, then human oversight, at least the simple checks and balances, will continue to be warranted.
Looking ahead, Malatsi’s department has a huge task.
Most importantly, “the policy should demonstrate verification. Sources should be checked carefully. Evidence should be traceable,” says Olivier. “Assumptions should be tested against real-world conditions. These steps may feel routine, yet they are the foundation on which public trust is built,” he says.
“If the next draft reflects careful oversight, disciplined verification and a clear understanding of local realities, this uncomfortable moment will have served a constructive purpose.”
Whether it’s addressing the trust deficit or minimising errors, the use of AI — especially where the stakes are so high — requires greater human intervention.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.