- South Africa's AI policy was pulled after at least six of its 67 academic citations, journals and authors no longer existed due to AI hallucinations.
- The policy itself was ambitious, proposing five new oversight bodies including a national AI commission and a road accident fund-style AI insurance superfund to compensate citizens harmed by AI.
- With no timeline for the revised draft, the scandal leaves South Africa's digital economy in regulatory limbo and serves as a stark warning to the continent about the use of unsupervised AI in government.
In a deeply ironic turn of events resonating throughout South Africa, Communications and Digital Technology Minister Solly Malatsi has withdrawn the draft of the country's national artificial intelligence policy after it was revealed that the document's bibliography contained fictitious sources. The scandal came to light when News24 reported that the policy document cited academic journal articles that do not exist, with several authors being credited with fundamental research who had never written on the topics attributed to them.
This mistake is especially stinging because this policy was designed to control the very technology that has destroyed it.
A policy on ghost research drafted
The draft policy was approved by the Cabinet on 25 March and 1 April, and published in the Government Gazette on 10 April 2026 for public comment, with submissions open until 10 June 2026. The highest levels of government, including President Cyril Ramaphosa's cabinet, approved before anyone could investigate whether its academic foundations were genuine.
Some of the 67 references listed in the draft either do not exist or point to articles not published in recognized journals. The editors of the South African Journal of Philosophy, AI & Society and the Journal of Ethics & Social Philosophy independently confirmed to News24 that their publications were never attributed there.
The diagnosis was almost immediate: AI hallucination. It appears that the drafters fed prompts into a generative AI tool, the same class of technology that was created to control the policy, and published the output without verifying a single quote.
the minister answers
In a statement posted to his X account on Sunday, Malatsi didn't mince words. “This failure is not merely a technical issue but has compromised the integrity and credibility of the draft policy,” he said. He said the Department of Communications and Digital Technologies “has not performed to the standard expected of an institution entrusted with the role of leading South Africa's digital policy environment.”
He said the most plausible explanation was that the AI-generated sources were included without proper verification, calling it an “unacceptable omission” that underscores “why vigilant human oversight over the use of artificial intelligence is critical.”
what is the policy proposed
Despite the way it was destroyed, the gist of South Africa's AI ambitions was actually broad. The draft AI policy proposes the establishment of a new AI governance ecosystem, including a National AI Commission, an AI Ethics Board, an AI Regulatory Authority, an AI Ombudsman, a National AI Safety Institute and a proposed AI Insurance Superfund, aimed at compensating individuals harmed by AI systems where liability is uncertain.
The AI Insurance Superfund was modeled on the Road Accident Fund, which was designed to compensate individuals or entities harmed by AI systems where liability is difficult to determine.
The policy also calls for integrating AI into school curriculum from primary to tertiary education, setting up community-based AI education centers in disadvantaged areas, and creating a labor market transition strategy to manage job displacement. On infrastructure, it proposed investments in supercomputing facilities, 5G and future 6G networks, high-capacity fiber and last-mile connectivity through low-Earth-orbit satellites, and went as far as framing universal internet access as a “socioeconomic right”.
The policy also seeks to align AI governance with POPIA with specific reference to automated decision making, and promote data security by design as a baseline requirement, along with mandatory watermarking of training data for large language models and cross-border data flow protocols to protect data sovereignty.
Importantly, the Department itself acknowledged the provisional nature of the document. An explanatory note published with the policy described it as a “work in progress” and said the government's final approach would require “extensive external consultation with both local and international experts and interest groups.”
political outcome
The scandal sparked intense inter-party political reaction. Khusela Diko, chair of parliament's communications committee, urged Malatsi to withdraw the draft amid credibility concerns, underscoring the political pressure surrounding the policy process.
Opposition politicians rejected the suggestion contained in Malatsi's response that responsibility could be placed on a junior official, arguing that the failure of due diligence lay with both the department and the ministry. The case was also compared to a recent scandal in which Deloitte was forced to return money to the Australian government over an AI-assisted report containing fabricated quotes.
Widespread implications for Africa
South Africa is one of the few African countries that has developed an AI policy, even as adoption of the technology continues to spread rapidly across the continent. The withdrawal means the country will now have to restart large parts of the consultation process, with no immediate timeline provided for the revised draft.
The episode also comes at an awkward moment globally. Governments around the world are racing to regulate AI, often relying on the same tools they are trying to control to help draft legislation and policy. South Africa's experience now stands as a cautionary reference point: the irony of AI policy confusing its academic foundations is not only shameful, it is a structural warning about the risks of deploying AI in high-risk governance tasks without rigorous human verification.
Minister Malatsi has promised that it will return with “much more rigorous oversight”. Countries and continents will be watching closely.
Statement on the Integrity of the Draft National Artificial Intelligence Policy
Following the revelation that the draft National Artificial Intelligence Policy published for public comment included various fictitious sources in its reference list, we began asking internal…
– SollyMalatsi (@SollyMalatsi) 26 April 2026
Source: CNBC Africa