Oversight, accountability, and AI moving into daily life

Writing in Business Day, Pitso Tsibolane argues that Cape Town's planning processes for data centres – the large facilities that house the computer servers powering AI and other digital services – are not transparent enough about the facilities' demands on electricity and water. The piece warns that approving data centres as though they were ordinary warehouses allows significant resource pressures to escape proper public scrutiny. For South Africa, where both energy supply and water security are already under strain, the argument raises questions about whether local government oversight is keeping pace with the infrastructure boom driven by growing demand for AI computing.

Discovery, the South African financial-services group, is taking a cautious approach to AI coding agents (software tools that write or modify code autonomously on a developer's behalf), according to TechCentral. The company is tracking the productivity gains these tools deliver while keeping tight limits on what they are allowed to do in its systems. The approach reflects a broader question facing South African businesses: how to capture the efficiency benefits of AI-assisted software development without losing control over critical infrastructure.

Writing in Business Day, Rufaro Mafinyani examines agentic AI – software systems that can take actions and make decisions on their own, without a human approving each step – asking whether this technology replaces workers or extends what they can do. The piece considers where such systems add genuine value, how organisations should build them responsibly, and who is accountable when an autonomous system acts in ways that cause harm. For South African businesses and policymakers beginning to adopt these tools, the accountability question is particularly live given the country's labour sensitivities and developing AI governance landscape.

Bolt, the ride-sharing app, has added support for OpenAI's ChatGPT as part of a global rollout that includes South Africa, according to Stuff South Africa. Riders who install a ChatGPT plugin can request fare estimates, compare routes, and check driver arrival times without opening the Bolt app directly, though bookings and payments still go through Bolt's own systems. The practical benefit is limited for now, and the source notes that Bolt's typical South African user base, which skews toward cost-conscious riders, may have little overlap with ChatGPT users.

Netcare, one of South Africa's largest private hospital groups, has submitted an AI-based clinical algorithm (a set of rules a computer uses to assist with medical decisions) for regulatory approval, according to ITWeb. The submission signals that AI tools are moving from pilot use into formal oversight processes in South African healthcare, with the South African Health Products Regulatory Authority likely to play a role in assessing the tool. How regulators evaluate and approve such systems will set a precedent for how AI is governed in clinical settings across the country.

Writing in Business Day, Duggan Matthews argues that as AI raises the general standard for analysis and information processing, distinctly human qualities such as creativity and judgement become more, not less, valuable in the workplace. The piece is aimed at a South African business audience weighing what skills and roles will matter as AI tools become more widely used.

Alibaba's Hong Kong-listed shares rose about 7% after the Chinese technology group unveiled Qwen3.8-Max, its largest AI model to date, with 2.4 trillion total parameters (the numerical settings a model learns during training) and strong early rankings on the Arena.AI benchmarking platform, according to Memeburn. The model uses a mixture-of-experts design, meaning only around 95 billion of those parameters are active at any one time, which can reduce computing costs and response delays. For South African developers and businesses, Memeburn notes that the broader competition between Chinese and US AI providers could improve pricing and customisation options, though practical questions around cloud access, regional latency, data governance, and the final licence terms are likely to matter more than the headline scale of the model.