South Africa has withdrawn its draft national AI policy after officials discovered at least six fabricated citations among the document’s 67 academic references. Communications Minister Solly Malatsi said the lapse compromised the policy’s credibility and promised consequences for those responsible. The draft — which proposed new AI oversight bodies and incentives for private investment — will be revised and reissued for public comment. The case highlights a rising trend of AI-generated "hallucinated" citations and the urgent need for strong human verification.
South Africa Withdraws Draft AI Policy After AI-Generated Fake Citations Exposed
South Africa withdraws AI policy after it was found written by AI
South Africa has withdrawn its draft national AI policy after officials discovered that portions of the document relied on AI-generated material that cited non-existent academic sources.
Communications Minister Solly Malatsi announced the withdrawal after investigators found that at least six of the policy’s 67 academic citations were fabricated outputs from an AI system. The fake citations pointed to articles that do not exist, even though the journals named are real publications.
"The most plausible explanation is that AI-generated citations were included without proper verification. This should not have happened," Mr Malatsi wrote on X. "This failure is not a mere technical issue but has compromised the integrity and credibility of the draft policy."
The draft had been published for public comment and aimed to position South Africa as a leader in AI innovation while addressing the ethical, social and economic challenges posed by AI technologies. It proposed the creation of new oversight bodies including a national AI commission, an AI ethics board and an AI regulatory authority, and outlined incentives such as tax breaks, grants and subsidies to spur private-sector investment in AI infrastructure.
How the Problem Came to Light
The issue was flagged after an investigation by News24 found that at least six citations listed in the document could not be verified. Editors at journals named in the draft — including the South African Journal of Philosophy, AI & Society and the Journal of Ethics and Social Philosophy — independently confirmed that the cited articles do not exist.
Minister of Communications and Digital Technologies of South Africa Solly Malatsi (R) speaks (AFP via Getty Images)
Why AI Produces Fabricated Citations
Experts say this episode illustrates a broader problem with generative AI: when a model lacks reliable information in a domain, it can produce confident-sounding but incorrect outputs, known as "hallucinations." Large language models such as OpenAI’s ChatGPT and Google’s Gemini are trained to predict likely word sequences, not to fact-check external sources, so they may invent plausible-looking references that are not real.
A study published in Nature reported that more than 2.5% of academic papers published in 2025 contained at least one potentially hallucinated citation, up from 0.3% in 2024. That increase translates to an estimated 110,000+ papers in 2025 with one or more invalid references.
Next Steps and Lessons
Government officials say the draft policy will be revised and reissued for public comment after corrections are made. The communications minister also warned that there would be consequences for those responsible for drafting the policy without adequate oversight.
This incident underscores the urgent need for rigorous human review of AI-assisted research and policy drafting. Institutions using generative AI must implement verification checks and accountability measures to prevent fabricated citations and protect the credibility of public documents.
Help us improve.
