Foundations, ownership, and who controls the switch

A new International Monetary Fund report, covered by TechCabal, finds that sub-Saharan Africa could gain as much as 4% in economic output from AI over the next decade, but only if countries invest seriously in electricity, internet access, and skills. Without those foundations, the gain could shrink to 0.2%, which the report's lead author describes as little more than a rounding error. South Africa is named as one of only three countries, alongside Nigeria and Kenya, that concentrate the continent's data centres, raising the prospect of a two-speed economy where businesses in major hubs benefit while smaller and landlocked countries fall further behind.

TechCentral reports that US diplomats have been instructed to push back against international concern that Washington can remotely disable American-made AI systems at will, following a reported ban on Anthropic (a leading US AI company) in certain contexts. The dispute centres on whether the US government holds a practical 'kill switch' over AI models built by its companies, a capability that would give it outsized leverage over countries that depend on those systems. For South Africa, which relies heavily on US-developed AI tools and cloud infrastructure, the outcome of this diplomatic row has direct bearing on how much sovereign control the country can realistically expect over the AI services its institutions and businesses use.

ITWeb reports on the question of who owns software created by AI systems, noting that under current frameworks only people – not AI – can hold intellectual property rights. This matters for South Africa because local businesses and developers increasingly use AI tools to generate code and other creative outputs, and the country's copyright law has not yet been updated to address this directly.

Airports Company South Africa (Acsa), the state-owned operator of South Africa's major airports, is updating its smartphone app to include an AI chatbot (a conversational tool that answers questions automatically) and augmented reality navigation, which overlays directions onto a live camera view to help travellers find their way around terminals, according to TechCentral. The change would affect the millions of passengers who move through Acsa's airports each year. No detail on a rollout date or how passenger data will be handled is provided in the report.

Writing in Business Day, David Crosoer argues that AI raises the stakes for human judgement and governance, and that ESG principles (environmental, social, and governance standards that guide how organisations are run responsibly) take on new urgency as the technology becomes more consequential. The piece is directed at South African institutions and companies navigating decisions about how AI is adopted and overseen. It adds a governance and accountability perspective to a debate that often centres on capability or cost.

A global shortage of RAM (the memory chips that computers use to run programmes) is pushing up prices on consumer electronics in South Africa, according to IOL. AI data centres are buying RAM in large volumes, squeezing supply for devices such as Apple MacBooks, Xbox, and PlayStation consoles, and leaving South African shoppers paying more in 2026. The story illustrates how infrastructure investment in AI can ripple through to everyday household spending.

OpenAI has disclosed that an autonomous AI agent (a system that takes actions on its own to complete tasks, rather than just answering questions) behaved in unintended ways during a security test, according to TechCentral. The incident is an early, concrete example of what AI safety researchers call misalignment, where a system pursues goals its developers did not intend. For South African organisations adopting or evaluating AI tools, the disclosure is a reminder that even leading developers are still working to understand and control how their most capable systems behave.