On April 2, South Africa’s Department of Communications and Digital Technologies (DCDT) released a draft artificial intelligence policy for public consultation. The proposal aimed to outline how the country plans to regulate and integrate AI across sectors, particularly in sensitive areas like finance and public services.
A central feature of the framework was its decentralized oversight model. Instead of placing responsibility under a single authority, the draft assigned various regulatory bodies to supervise different aspects of AI development and deployment. It also introduced a risk-based classification system, categorizing AI applications as “unacceptable,” “high,” “limited,” or “minimal” risk—an approach intended to guide adoption and safeguard the public.
However, what began as a forward-looking initiative quickly turned controversial.
Fabricated Citations Raise Red Flags
Critics soon identified serious flaws in the document. Reports revealed that at least six of the references cited in the draft policy could not be verified. Some sources appeared nonexistent, others could not be traced to any recognized academic publication, and a few were suspected to be entirely fabricated—potentially generated by AI itself.
This discovery has fueled criticism and raised concerns about the integrity of the policy-making process. The irony has not been lost on observers: a document designed to regulate AI risks may have itself relied on unverified AI-generated content.
The issue has drawn attention from both media and political figures, intensifying scrutiny on the DCDT.
Political Pressure Mounts
The controversy quickly escalated into a political issue. Lawmakers, including Khusela Diko, Chairperson of the Portfolio Committee on Communications, called for the immediate withdrawal of the draft. Questions have also been directed at Solly Malatsi, the Minister responsible for the department.
In response, Malatsi acknowledged the concerns and announced an internal investigation. He instructed the department’s Director General to identify how the errors occurred and hold those responsible accountable.
By Sunday, Malatsi publicly confirmed that the problematic citations were indeed present and announced the withdrawal of the draft policy. He suggested that the most likely explanation was the inclusion of AI-generated references without proper verification—a lapse he admitted should never have happened.
Government Scrambles to Contain Fallout
The incident has placed South Africa’s government in a difficult position. What was intended to demonstrate leadership in AI governance has instead exposed weaknesses in oversight and quality control.
Officials are now working to manage the reputational damage while reassessing the policy framework. Despite the setback, the draft did offer a structured approach to addressing AI-related risks, signaling a broader awareness of the technology’s growing role in public systems.
Regional Context Highlights Different Approaches
South Africa’s attempt at a decentralized regulatory model stands in contrast to strategies in other African countries. For instance, Nigeria has leaned toward a more centralized approach, with institutions like the central bank encouraging financial institutions to adopt AI tools for fraud detection and anti-money laundering efforts.
Similarly, Kenya has pursued frameworks that consolidate oversight under fewer authorities, aiming for streamlined governance.
A Cautionary Moment for AI Policy Development
The episode serves as a cautionary tale about the risks of relying on AI without rigorous human oversight—especially in official policy documents. While the ambition to regulate emerging technologies is commendable, the execution has underscored the importance of credibility, verification, and accountability.
As South Africa revisits its AI strategy, the focus will likely shift toward rebuilding trust and ensuring that future policies meet the standards they seek to enforce.
