South Africa withdraws AI policy draft over fake citations scandal

South Africa pulled its first national artificial intelligence policy draft after officials found that the reference list contained fictitious sources that appeared to have been generated by AI, a failure that turned the government’s own technology roadmap into a test of credibility. Communications and Digital Technologies Minister Solly Malatsi said the most plausible explanation was that AI-generated citations had been inserted without proper verification, and he said the mistake compromised the policy’s integrity and credibility.

The draft had been approved by Cabinet on March 25 and again at a special sitting on April 1, then published in the Government Gazette on April 10 for public comment, with submissions due by June 10 at 4 p.m. It was meant to set South Africa’s ambitions as a continental AI leader while also addressing ethical, social and economic concerns. The document proposed new institutions, including a National AI Commission, an AI Ethics Board and an AI Regulatory Authority, and it also contemplated incentives such as tax breaks, grants and subsidies to support private-sector collaboration.

Malatsi cast the withdrawal as more than a paperwork problem. He said the department had not met the standard expected of an institution entrusted with leading South Africa’s digital policy environment, and he signaled that there would be consequence management for those responsible for drafting and quality assurance. He also said the lapse showed why vigilant human oversight remains essential when governments use artificial intelligence in drafting, research and public administration. A revised version will come later, but no timetable has been given.

The embarrassment lands at a sensitive moment for a government trying to present itself as serious about AI governance. The draft was built around a phased approach and framed AI as a tool for responsible innovation, job creation and better access to skills, while also embedding intergenerational equity so that benefits would not be captured by one generation alone. Instead, the episode has exposed a basic institutional weakness: South Africa moved quickly to regulate AI before proving that its own review process could reliably verify the sources behind the policy.