South Africa Withdraws National AI Policy After Fabricated Citations Scandal

South Africa's Department of Communications and Digital Technologies has withdrawn its draft National Artificial Intelligence Policy from the public review process after journalists discovered that a significant portion of the document's academic bibliography did not exist.

Minister Solly Malatsi announced the withdrawal on 27 April 2026, days after an investigation by News24 found that at least six of the 67 scholarly citations underpinning the policy's evidence base pointed to fabricated journals, non-existent papers, or misattributed authors — the hallmark pattern of unverified generative AI output rather than conventional research error.

The department confirmed that two officials involved in drafting the document have been placed on precautionary suspension while an internal inquiry determines how the citations entered the policy and whether appropriate verification procedures were bypassed. Malatsi said the government would not proceed with a policy meant to govern the responsible use of artificial intelligence while its own drafting process stood accused of using the technology carelessly.

### A policy built to guide AI, undone by AI

The draft policy had been positioned as South Africa's foundational framework for AI governance — covering data protection, algorithmic accountability, sector-specific guidance for finance and healthcare, and the government's approach to attracting AI infrastructure investment. Civil society groups and academic reviewers who had flagged inconsistencies during the public comment period said the episode underscores a broader risk: policy documents assembled with AI drafting tools, without rigorous citation verification, can quietly launder synthetic claims into binding public policy.

The department has not given a firm timeline for resubmitting a revised policy, saying only that future drafts will undergo an expanded verification process before publication. The withdrawal leaves South Africa — one of the continent's most active AI investment destinations — without a finalised national AI policy at a moment when data centre capacity, GPU access, and AI-sector employment are all expanding rapidly across the country.

The incident has drawn comparisons internationally to earlier cases of AI-hallucinated citations surfacing in legal filings and academic work, but officials say this is among the first instances of a national government policy document being withdrawn for the same reason.