South Africa withdraws AI policy after discovering it cited fake AI-generated research

South Africa's AI Policy Faces Embarrassing Withdrawal

South Africa's ambitious attempt to position itself as a continental leader in artificial intelligence came to an abrupt halt when the country's draft national AI policy was withdrawn just 16 days after its official publication 1

1

. On April 10, the Department of Communications and Digital Technologies published the Draft South Africa National Artificial Intelligence Policy for public comment, but journalists quickly discovered that the document contained fabricated references—a telltale sign of AI hallucinations 3

3

.

The fictitious research citations fell into two categories: academic journals that do not exist, and real journals in which the referenced research articles were never published 1

1

. News24, a local news website, reported that at least six references in the document were fabricated, with experts confirming that the errors matched classic generative AI problems—convincing on the surface, entirely made up underneath 3

3

.

Minister Acknowledges Failure of Human Oversight

Communications Minister Solly Malatsi was frank in his assessment of what went wrong. "This failure is not a mere technical issue but has compromised the integrity and credibility of the draft policy," he wrote in a post on X 4

4

. The minister acknowledged that the most plausible explanation was that fake AI-generated sources were included without proper verification, stating emphatically: "This should not have happened" 2

2

.

Malatsi emphasized that this unacceptable lapse proves why vigilant human oversight over the use of artificial intelligence is critical, adding that it was "a lesson we take with humility" 5

5

. The minister indicated that those involved in drafting and sign-off can expect "consequence management," though he did not specify when a revised policy would be released 3

3

.

What South Africa's AI Policy Promised

The withdrawn policy had outlined ambitious plans to establish new institutions, including a National AI Commission, an AI Ethics Board, and an AI Regulatory Authority 4

4

. It also proposed creating incentives such as tax breaks, grants, and subsidies to encourage private-sector collaboration and stimulate investment in AI innovation 5

5

.

The document was sold as a forward-looking framework built on principles of responsible AI governance, including accountability, transparency, and explainability—non-negotiable conditions echoed by the Organisation for Economic Co-operation and Development principles and the Smart Africa AI Blueprint 1

1

. Ironically, South Africa's AI policy failed to meet all three of these standards in its own production.

Critical Questions About Transparency and Accountability

The compromised policy integrity raises serious questions about transparency and accountability in AI in governmental processes. The Department of Communications and Digital Technologies has not disclosed which sections of the policy are materially affected by the fabricated sources, which chatbot or generative AI tool was used, by whom, or at which stage of drafting the hallucinated citations entered the document 1

1

.

The hallucinated sources did not just invent references—they manufactured seemingly credible African scholarly authority, casting highly respected authors' names in a false light and attributing false evidence to real institutions recognized as authoritative publishers of academic papers 1

1

. This undermines both information integrity and credibility, the public's reasonable expectation that information from an authoritative source can be trusted.

A Broader Pattern Emerges

This incident is not isolated within South Africa's government. Just four days after the policy withdrawal, Home Affairs Minister Leon Schreiber suspended two senior officials after references in a cabinet-approved policy document on immigration were also flagged as AI hallucinations 2

2

. Both ministers belong to the Democratic Alliance party, which has championed the adoption of modern technology to boost government efficiency since joining the ruling coalition as the second-biggest party in 2024 2

2

.

Khusela Sangoni-Diko, chair of the parliamentary portfolio committee overseeing the department, publicly told Malatsi to pull the document before it caused further embarrassment, suggesting that the redraft skip "using ChatGPT this time" and that the government should stop looking for a scapegoat, or "scape-bot" 3

3

.

Global Implications for AI Governance

Phony citations generated by AI have become a particular problem in legal documents worldwide. An online legal hallucination database maintained by lawyer and data scientist Damien Charlotin has found more than 900 such cases in the United States alone, with four previously documented in South Africa 5

5

. A growing number of U.S. lawyers have been reprimanded for submitting AI-generated legal briefs riddled with hallucinations, while in Australia, Deloitte had to help clean up a government report after AI-generated citations and even a made-up court quote slipped through 3

3

.

The South African case underscores that the harms of generative AI extend beyond fake references to include fake images, fake videos, fake voices, and the weaponization of people's likenesses through deepfakes and disinformation 1

1

. When a national AI policy cannot tell real sources from imaginary ones, it raises fundamental questions about readiness to regulate AI governance more broadly 3

3

.