South Africa's Redrafted AI Policy Delayed Until 2027 Amid Drafting Challenges and Shifting Timelines

South Africa's updated national artificial intelligence policy is not expected to be released for public comment until at least March 2027, according to an official from the Department of Communications and Digital Technologies. This timeline represents a significant delay from previous commitments, pushing back public engagement on a critical policy document that will shape the country's AI landscape.

The delay follows a previous attempt to gazette a draft policy in April, which was quickly withdrawn due to the inclusion of fictitious sources in its reference list. In response, Minister Solly Malatsi appointed an independent expert panel in May, chaired by Wits University AI researcher Benjamin Rosman, to advise on the redrafting process. A government drafting team, including representatives from the Presidency and other key departments, is also involved to ensure robustness.

The current focus of the policy discussions includes crucial aspects like compute and data sovereignty, as well as "model sovereignty" to ensure large language models can effectively handle South Africa's diverse official languages. However, a significant gap identified is the urgent need for investment in computing infrastructure to support these ambitions. An implementation roadmap is being developed concurrently to facilitate immediate action once the policy is finalized.

Despite South Africa's perceived regional lead in AI readiness, the official noted that other African nations, such as Egypt with its Karnak national LLM, are demonstrating faster progress in AI implementation. The protracted policy development process could impact South Africa's pace in leveraging AI for national development and maintaining its regional standing.