• Draft AI policy pulled after fictitious references were discovered

• Authorities say unverified AI-generated citations likely caused the issue

• Incident raises concerns over governance and use of AI in public policy

South Africa has withdrawn its draft national artificial intelligence policy after discovering that the document included fabricated references, likely generated by AI tools.

The decision was announced on April 26 in a post on X by Communications and Digital Technologies Minister Solly Malatsi, who acknowledged a breakdown in the government’s validation process.

“The Department of Communications and Digital Technologies did not deliver on the standard that is acceptable for an institution entrusted with the role to lead South Africa’s digital policy environment,” the minister said. “The most plausible explanation is that AI-generated citations were included without proper verification”.

The draft, which had been released for public consultation, contained academic citations and sources that do not exist. These issues came to light during a detailed review, revealing that some content had not been properly checked. Authorities decided to withdraw the document, citing concerns over its credibility.

The episode highlights a striking irony: a policy meant to regulate artificial intelligence was itself undermined by improper use of the technology. It also underscores the limits of generative AI tools, which can produce convincing but inaccurate content without careful human oversight.

Beyond the immediate incident, the issue raises broader institutional concerns. Public policy—especially in areas as sensitive as digital governance—relies on accurate and verifiable information. The inclusion of false sources calls into question internal review processes and how public institutions are integrating AI into their workflows.

South Africa remains one of the most advanced countries in sub-Saharan Africa in terms of AI readiness. According to the AI Readiness Index published by UK-based Oxford Insights, the country scored 53.94 out of 100, ahead of Mauritius (52.76) and Kenya (52.55). The continental average stood at 29.12, reflecting significant gaps in infrastructure, talent development, and regulatory frameworks.

More broadly, the incident may serve as a warning. As governments accelerate the adoption of AI, it underscores the need for strong safeguards to ensure that the technology does not undermine the credibility of the very institutions seeking to regulate it.

Adoni Conrad Quenum