The Department of Communications and Digital Technologies has placed two officials on precautionary suspension after the now-withdrawn draft national artificial intelligence policy was found to contain fake AI-generated references, deepening scrutiny over how government departments are using generative tools in policy work. The move comes just days after a similar scandal at home affairs, making this the second South African department in less than a week to act over AI hallucinations embedded in a cabinet-approved document.
The suspensions were confirmed over the weekend by director-general Nonkqubela Jordan-Dyani, who said the action takes effect immediately while an investigation continues into how the draft came to include fabricated academic citations. Those citations were exposed after News24 reported that several of the sources did not exist, while others were attributed to real journals and researchers who had never written on the claims cited in the document.
Jordan-Dyani said the department had already launched an internal review to establish what happened and who was responsible for the drafting and quality-control failures. “The irresponsible use of AI tools compromised the integrity of the policy document,” she said, adding that the step was meant to reinforce accountability within the department. The two officials have not been publicly named.
The controversy has put a sharp spotlight on a document that had already been through the formal machinery of government. The draft policy was approved by Cabinet on 25 March and later gazetted on 10 April for public comment before it was ultimately withdrawn. That sequence has raised uncomfortable questions about how a policy paper containing bogus references made it through official checks in the first place.
Communications minister Solly Malatsi pulled the draft policy last Sunday and said those involved in both drafting and quality assurance would face “consequence management”. He described the lapse as “unacceptable”, and said the incident showed why human oversight over the use of AI remains essential, even as government departments explore the technology’s potential benefits.
The issue is bigger than one department. Across the public sector, the rapid adoption of generative AI tools has created new risks around verification, source checking and document integrity. In this case, the problem was not simply a typo or a formatting error. It involved invented references presented as credible academic support in a policy that had already received Cabinet approval.
For ordinary South Africans, the scandal will likely feel familiar: another example of the state being asked to modernise while basic controls appear to have fallen away. AI can speed up research and drafting, but it can also produce convincing nonsense if human editors do not catch errors before publication. In policymaking, where accuracy matters and every citation can shape debate, that risk is especially serious.
AI hallucinations in a draft policy raise fresh questions about South African government AI policy
The communications department’s action follows a similar move by the Department of Home Affairs, which last Thursday suspended two officials after AI hallucinations were discovered in the reference list attached to the revised white paper on citizenship, immigration and refugee protection. That document too had gone through high-level approval before the errors were detected.
The near-simultaneous scandals have prompted fresh concern about the spread of unchecked AI use inside government. If two separate departments could submit official policy material containing fabricated references, then the larger question is not just who is responsible in these cases, but whether clear rules exist at all for the use of generative AI in the state sector.
As we reported earlier, the concern is not only reputational. When policy documents rely on false citations, the credibility of the entire process is undermined. Parliament, Cabinet, civil society and the public all depend on the assumption that government documents are properly researched and reviewed. Once that trust is damaged, the fallout can spread well beyond one department or one minister.
The timing is also significant. South Africa is in the middle of a broader policy conversation about digital transformation, regulation and the role of AI in the economy. The draft national artificial intelligence policy was meant to help frame that debate. Instead, it has become a cautionary tale about what happens when the technology is used without strict editorial discipline and clear accountability.
Our sources indicate that the internal review inside the communications department will now have to establish not only who used AI tools, but also whether supervision was absent at more than one level. That means tracing the drafting process, checking whether the references were verified, and determining where the quality-control chain broke down.
There is also the issue of public trust. Once a policy paper has been exposed as containing fabricated material, any future version will be read through a lens of suspicion. Even if the substance of the policy is restored, the department will now have to work harder to convince stakeholders that the process is credible and that the mistakes were not systemic.
At this stage, the department has not said when the investigation will be completed, and it remains unclear whether further disciplinary action will follow. What is clear is that the fallout from the fake AI references has moved beyond embarrassment. It has become a test case for how South Africa’s public sector manages the use of emerging technology in official decision-making.
For now, the message from both communications and home affairs is the same: government can no longer treat AI-generated material as if it is automatically reliable. The technology may be powerful, but the responsibility for accuracy still sits squarely with people. And as these back-to-back suspensions show, that lesson is being learned the hard way.