Communications minister Solly Malatsi is under intensifying pressure after the ANC study group on communications and digital technologies demanded that he explain how South Africa’s draft national AI policy ended up containing what appear to be AI-generated hallucinations. The controversy has now moved well beyond an embarrassing drafting error and into the realm of parliamentary accountability, with questions being raised about oversight, quality control and whether the department is equipped to guide the country’s AI future.
At the centre of the storm is a policy meant to set the rules for one of the most fast-moving technologies in the world. Instead, the draft has been pulled after it emerged that several references in the document were either fake, unverifiable or linked to journals that could not confirm the cited work existed. That has triggered a political backlash, a public credibility crisis and renewed scrutiny of how government handles technology policy in the age of artificial intelligence.
The ANC study group said late on Tuesday that Malatsi should be called before parliament’s portfolio committee on communications at the earliest available opportunity. In plain terms, they want him to explain how a policy designed to govern AI was itself apparently drafted with the help of AI, and why nobody caught the problem before it was published for public comment.
“This debacle represents one of the most alarming failures of ministerial oversight and intellectual rigour in the recent history of South Africa’s digital governance,” said Imran Subrathie, the study group’s chief whip. Subrathie, who also sits on the portfolio committee, said the department’s handling of the matter showed a troubling lack of discipline at exactly the moment South Africa needs clarity and competence on emerging technologies.
He added that the group is “deeply troubled” that a document meant to shape the country’s response to what many call the defining technology of the century was produced in this way. For the ANC MPs now circling the issue, the embarrassment is not just about a few bad citations. It is about whether the department can be trusted to produce a credible AI framework at all.
The study group itself cannot summon Malatsi directly. That power lies with the portfolio committee, which can compel attendance under section 56 of the Constitution. The committee is chaired by Khusela Diko, who has already publicly called on the minister to withdraw the draft and go back to the drawing board.
AI policy controversy puts Solly Malatsi under parliamentary fire
What started as a policy drafting scandal has quickly turned into a wider political battle. The ANC study group wants Solly Malatsi to give a full account of the internal review process, identify the officials responsible for drafting and quality assurance, and explain what consequence management he intends to implement. It also wants the department to commit to a redraft driven by human expertise, proper evidence and credible oversight rather than what it described as “AI-generated shortcuts”.
Malatsi, who is one of the more senior DA figures serving in the government of national unity, withdrew the draft on Sunday evening after an internal review he ordered found that parts of the reference list included fictitious sources. The minister said the likely explanation was the use of AI-generated citations that had not been properly checked, and he promised that there would be consequence management for those involved.
That admission only deepened the political fallout. The draft policy was an 86-page document with 67 references, but several of those citations led nowhere. Some referred to journals that had no record of the article in question, while others pointed to publications that simply did not exist. Editors from the South African Journal of Philosophy, AI & Society and the Journal of Ethics and Social Philosophy later confirmed that the articles attributed to them were never published there.
The public comment period had been set to close on 10 June, but with the policy now withdrawn, it is still unclear when a revised version will appear. That uncertainty matters, because South Africa has been moving slowly but steadily into the AI policy space, and this setback risks delaying the country’s attempt to catch up with global developments.
Tuesday’s intervention from the ANC has sharpened the political tone around the issue. Diko, who effectively serves as Malatsi’s most senior parliamentary overseer on the matter, had already taken to X to say the draft should be scrapped and redrafted “without using ChatGPT this time”. She also accused the minister of looking for a “scape-bot” in the aftermath of the scandal.
That remark was not left hanging. Dean Macpherson, the DA’s public works and infrastructure minister, jumped in to defend Malatsi on the same platform, dismissing Diko’s criticism as grandstanding. Diko hit back, saying DA ministers embodied populism and would soon find out that governance is harder than it appears from the outside.
Malatsi’s position is politically sensitive for another reason too: he is the first non-ANC minister to head the communications portfolio. That has made him a focal point in the often tense relationship between the GNU partners and their parliamentary counterparts. Disputes over broad-based black economic empowerment in the ICT sector and Starlink’s pursuit of a South African operating licence have already placed him in difficult territory.
The AI policy debacle lands on top of all that. For critics, it is not an isolated administrative mistake but another sign that the department is struggling to navigate the policy demands of a digital economy that is moving faster than government structures can comfortably handle. For Malatsi, that means the question is no longer only about a bad draft. It is about whether his department can credibly lead national AI regulation at all.
The policy itself had already attracted criticism before the reference scandal broke. Technology investor Stafford Masie argued in an open letter last week that the draft risked regulating South Africa out of the global AI economy by focusing too heavily on governance structures instead of investment in infrastructure. He pointed out that the document proposed an ambitious new framework of seven institutions to oversee AI, including a National AI Commission, an AI Ethics Board and an AI Safety Institute, even though government had not committed funding for the compute infrastructure such bodies would likely need.
That criticism matters because it speaks to a broader policy dilemma facing South Africa. The country wants to build guardrails for AI, but it also needs to support innovation, investment and practical implementation. A policy that overreaches without the resources to back it up risks becoming symbolic rather than effective.
Former DA MP and digital rights commentator Phumzile van Damme has also weighed in, describing the episode as yet another example of government’s technology illiteracy. She warned that the saga could become a global cautionary tale about AI misuse, especially given the irony of a policy on artificial intelligence being tripped up by artificial intelligence itself.
Van Damme also pointed to the Deloitte Australia case, where the consultancy was forced to refund part of a government fee after a report was found to contain fabricated academic references. Her point is clear: this is not just a South African embarrassment, but part of a growing international conversation about how easily AI tools can produce convincing nonsense when they are used carelessly.
For Malatsi, the damage is already done. The department now has to rebuild trust in the process, reassure parliament that there was proper oversight, and produce a revised policy that can survive public and expert scrutiny. Until then, the story will continue to dog the minister and his department.
As we reported earlier, the issue has now shifted from a technical drafting error to a test of political accountability. If parliament decides to act, Solly Malatsi may soon have to answer difficult questions in public about who drafted the policy, who signed off on it, and how South Africa’s AI roadmap was allowed to be derailed by the very technology it was meant to regulate.