South Africa Withdraws AI Policy After Finding It Contains Fake Citations
South Africa's draft national AI policy has been withdrawn from parliament after officials discovered it included fabricated citations, apparently generated by artificial intelligence.
The document was nearing finalization when the fake references surfaced in its source list. Solly Malatsi, South Africa's Minister of Communications and Digital Technologies, announced the withdrawal on April 26, saying the lapse compromised the policy's credibility.
"This failure is not a mere technical issue but has compromised the integrity and credibility of the draft policy," Malatsi said. "The most plausible explanation is that AI-generated citations were included without proper verification. This should not have happened. In fact, this unacceptable lapse proves why vigilant human oversight over the use of artificial intelligence is critical."
A recurring problem across industries
Hallucinated citations have become a persistent issue with generative AI and large language models. Language models frequently produce plausible-sounding but entirely false references, a problem that has proven difficult to solve.
Legal documents have been hit particularly hard. U.S. lawyers have faced discipline and reprimands for submitting AI-generated briefs filled with hallucinations. A database tracking such cases found more than 900 instances in the United States alone, with four documented in South Africa prior to this policy incident.
What the policy aimed to accomplish
The withdrawn draft outlined establishing a national AI commission, an ethics board, and a regulatory body. It also included tax incentives, grants, and subsidies designed to attract private-sector investment.
South Africa's goal was to position itself as Africa's leading hub for AI innovation. Malatsi did not provide a timeline for when a revised draft would be ready.
For writers and other professionals using AI tools, the incident underscores a practical concern: AI for writers and other applications requires human verification at every stage. Trusting AI output without checking facts and sources remains risky.
Your membership also unlocks:
