🚨 SA's AI Policy Pulled: The Hallucination That Fooled Cabinet

This is the ultimate cautionary tale for our community. In a case of peak 'hallucinatory irony', South Africa's Draft National AI Policy -- the document meant to regulate this tech -- was withdrawn this week after it was caught making things up. 📉

Here's the breakdown:

• The Fiasco: On Sunday, April 26, Minister Solly Malatsi withdrew the 86-page draft just weeks after it was gazetted for public comment.
• The Smoking Gun: Investigative reporting by News24 found at least six fictitious academic citations in the bibliography. The "papers" looked real and cited legitimate journals, but they do not exist—a textbook case of unchecked LLM output. 🤖🤥
• The Failure: The document had already been approved by Cabinet. It passed through multiple high-level "quality assurance" layers without anyone clicking a link to verify the research.
• The Fallout: Malatsi called it an "unacceptable lapse," admitting that AI-generated citations were likely included without verification. He has promised "consequence management" for those responsible. ⚖️

The Lesson for Prompt Crafters
1. Verification is the only vaccine: AI is a linguistic engine, not a fact-checker. Never trust a citation, quote, or legal reference without manual verification. 🔍
2. Formatting != Accuracy: Fake sources can look incredibly professional. The more "academic" an AI looks, the more vigilant you need to be.
3. Human-in-the-Loop is non-negotiable: If you use AI to write the rules for AI, and the AI hallucinations the rules, the system fails. Our value as journalists is the "verification layer" that the DCDT skipped. ✍️✅

Sources: DCDT Ministerial Statement / News24 / TechCentral (April 26–30, 2026).