The AI Hallucination Hazard: Lessons from South Africa’s AI Policy Mistake

In the world of ICT, we often talk about “garbage in, garbage out.” Usually, we are referring to data processing. Last week, the South African Department of Communications and Digital Technologies (DCDT) provided a literal example of this at the highest level of policy.

Minister Solly Malatsi took the unprecedented step of withdrawing the draft National AI Policy after it emerged that sections of the document contained hallucinated citations and fake research. This was not just a technical glitch. It was a failure of process that highlights the risks of using AI without human oversight.

I have spent over 25 years navigating IT governance and strategy. If there is one thing I have learned, it is that technology should enable human expertise, not replace it.

Minister Malatsi’s response was swift. He took full responsibility, stating he was embarrassed and that the error should not have happened. In a political landscape where “the dog ate my homework” is a common defense, this level of accountability is refreshing.

However, an apology is only the beginning. For a document as critical as a national AI framework to contain fake data, there was a clear breakdown in internal quality assurance. We need to see proper consequence management within the DCDT. Policies that shape the future of our economy cannot be “copy-pasted” or generated by a chatbot without rigorous verification.

Reports suggest that local South African AI experts were largely excluded from the drafting process. This is the real tragedy. We have a wealth of brilliant minds in our universities, tech hubs, and private sector who understand the local context.

Government must foster and consult industry experts early and often. When you sideline the people who actually build and study these systems, you end up with a policy that lacks credibility. Strategic ICT delivery requires a “sleeves-rolled-up” approach, not a boardroom-only perspective.

While we fix our internal processes, the rest of the world is moving at light speed. We cannot afford to be left behind in the AI arms race.

• The United Arab Emirates was the first to appoint a Minister of State for Artificial Intelligence in 2017.
• The European Union has already passed the EU AI Act, creating a comprehensive legal framework for AI governance.
• China and the US are investing billions into sovereign AI capabilities to ensure national security and economic dominance.

These countries are not just writing papers. They are building dedicated departments and cabinets to manage the ethical and economic impact of AI.

South Africa needs to accelerate this policy with a new sense of urgency and precision. We need an AI framework that serves our specific needs, addressing unemployment, improving education, and securing our digital borders.

Technology is an enabler, not noise. We must use it strategically to align with business and national value. We need a policy drafted by humans, supported by data, and validated by experts.

What is your take on the Minister’s decision to withdraw the policy? Should we be looking at a dedicated AI Ministry to prevent these errors?

Let’s start the conversation → Contact www.m-konsult.com/contact or connect with me on LinkedIn

Other articles that may interest you: https://m-konsult.com/news/