{
  "id": 13105715,
  "title": "From Users to Builders: The Test Facing South Africa’s Universities",
  "url": "https://urgent.news/2026/10/09/from-users-to-builders-the-test-facing-south-africas-universities",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-09T11:44:00.000Z",
  "source": {
    "name": "IOL",
    "slug": "iol",
    "url": "https://iol.co.za/news/brics/2026-10-09-from-users-to-builders-the-test-facing-south-africas-universities/"
  },
  "original_language": "en",
  "account": "From Users to Builders: The Challenge for South Africa’s Universities\n\nManamela stressed the urgency of universities adapting to AI, questioning arguments, testing evidence and accepting responsibility for knowledge. However, AI can create content without verifying its truth. In the job market, this is significant; in the second quarter of 2026, those with a matric certificate faced a 12.4% unemployment rate, while the rate for the unemployed without matric was 39.2%, and youth unemployment stood at 47.4%. Yet, this protection is not certain; in the first quarter, unemployment among those with below-matric education decreased year-on-year, but increased among those with matric and graduates.\n\nThe government approved the draft National AI Policy on 25th March 2026, with public comment open until 10th June, and full implementation expected by 2027 or 2028. The policy's six pillars include capacity and talent development, but there is a risk that these governance structures arrive before funded programs. Regarding language data, none of the top 34 languages on the internet are African, and progress is primarily driven by communities and donors. The Africa Next Voices initiative recorded 9,000 hours of speech across 18 languages, including isiZulu and Tshivenda, while the Masakhane hub, funded by Google.org, the FCDO, IDRC and the Gates Foundation, is developing datasets for 50 languages. However, there seems to be a lack of sustained public funding.\n\nCapacity is arriving through private operators, such as Stratos Lab deploying 400 Nvidia B300 GPUs at a Centurion data centre and Cassava taking delivery of 3,000 GPUs in South Africa. Nevertheless, Africa accounts for less than 1% of global data centre capacity. This dependence is also evident at the research level, where Lelapa AI obtained compute credits from Microsoft's AI for Good Lab to train its InkubaLM model.\n\nFor universities, the terms of access to these resources matter as much as their availability. Manamela urged universities to move beyond theoretical discussions and make practical decisions about teaching, assessment, and research capacity. Three areas appear most important: technological change will require people to continuously learn throughout their working lives, while qualifications must maintain depth and standards. Modular, stackable credentials could be a practical solution. The policy sets the right test, but it is not enough to fund GPUs or datasets. Four factors will determine whether it leads to increased capacity: whether the government moves past theory and implements practical actions, whether sufficient resources are allocated, whether there is a focus on practical skills development, and whether universities can adapt to the rapid pace of technological change.",
  "summary": "At the 4th Universities South Africa (USAf) conference this week, Higher Education and Training Minister Buti Manamela set universities a two-part task. They must produce graduates who can use AI tools intelligently and question them competently, and researchers capable of helping to build the technology. The first part is largely within a university’s control. The second depends on policy, data…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}