Two South African Home Affairs Officials Suspended Over AI Hallucinations in Policy Paper

The Department of Home Affairs (DHA) in South Africa suspended two officials after discovering AI-generated hallucinations in the reference list of a revised white paper on citizenship, immigration, and refugee protection. The suspensions affect the Chief Director of the citizenship and immigration unit and the director involved in drafting the document.

What Happened

The discrepancies were found in the reference list attached to the white paper. The references were deemed to be hallucinations — erroneous or fictitious outputs from large language models (LLMs). According to the DHA statement, the references appear to have been generated and attached after the fact, as they are not cited in the body of the text.

Response and New Procedures

The DHA acknowledged the embarrassment and said it will use the incident to modernize processes. Moving forward, they will design and implement AI checks and declarations as part of internal approval processes. Two independent law firms have been appointed to manage the disciplinary process and review all policy documents produced since 30 November 2022 — the date ChatGPT was released for public use.

The DHA maintains that the revised policy accurately reflects the government's position and stands by its contents, stating the hallucinations were confined to the standalone reference list.

Broader Context

This incident follows a similar one a week earlier, where the Department of Communications and Digital Technologies (DCDT) withdrew its draft National AI Policy after fictitious sources were found. Minister Solly Malatsi noted: “The most plausible explanation is that AI-generated citations were included without proper verification.”

The DHA accepted AI's growing use and said institutions must adapt: “It is a transformative but disruptive technology that is changing how organisations operate across the private and public sectors. We must now adapt to keep up.”

This case highlights a real-world consequence of using LLMs for document drafting without rigorous verification — especially in government where accuracy is critical. For developers working with AI agents, it reinforces the need for validation layers, citation checks, and human-in-the-loop review.