Inequality, fraud, and data rights in an AI-shaped economy

South African and global insurers are facing a new wave of fraudulent claims built on AI-generated images, where tools that can fabricate convincing photographs of damage or accidents are now widely accessible, according to IT News Africa. The concern is that this lowers the skill and cost required to commit insurance fraud, potentially shifting the volume and nature of false claims that companies must detect. For South African consumers, the risk is indirect but real: fraud losses are typically built into premium pricing, so a rise in successful claims of this kind can push costs up for everyone.

Kandua, a home services platform owned by South African insurer Santam, has launched a conversational AI tool called Jess that is designed to guide homeowners through finding and managing repair and maintenance work. According to IT News Africa, the company describes Jess as the first AI home companion of its kind in South Africa, with a stated aim of improving trust and safety in a sector where consumers often struggle to verify the reliability of service providers. The launch is a local example of AI being applied to a consumer services market, though independent assessment of the product's claims has not been reported.

Writing in Business Day, South African commentator Heath Muchena argues that the build-out of AI infrastructure (the physical and software systems needed to run AI at scale) is reshaping how investors assess risk, alongside shifts toward digital settlement and tokenised assets (financial instruments recorded on a digital ledger). The piece contends that these converging changes make non-sovereign stores of value, such as Bitcoin, more relevant to portfolio thinking. The argument is relevant to South African investors navigating a financial system that is beginning to absorb both AI-driven processes and new forms of digital money.

A University of the Witwatersrand professor of information systems, writing in The Conversation, argues that the spread of AI-powered chatbots and automated service tools risks creating what he calls 'relational apartheid': a pattern in which access to genuine human attention becomes unevenly distributed, with wealthier or better-connected people still dealt with by humans while others are routed through machines. Drawing on ubuntu, a southern African philosophy that understands personhood as formed through mutual relationships rather than in isolation, he contends that simulated responsiveness is not the same as real recognition, and that in a society the World Bank ranks among the world's most unequal, existing gaps in income, language, and institutional access could reappear in digital form. He calls for AI to be presented clearly as a tool rather than a companion, and for it to support human professionals in sensitive settings such as healthcare and education rather than replace them.

Google has begun using users' search history to train its generative AI systems (software that produces text, images, or other content from patterns learned in large datasets), according to Stuff South Africa. Users were notified by email rather than asked for consent, but can opt out by visiting My Google Activity and switching off Search Services History – though Google's own interface notes that data already used for training may be retained for up to four years even after deletion. For South Africans, the piece raises a practical question about data rights: the country's privacy law, the Protection of Personal Information Act (POPIA), sets rules on how personal data may be processed, and whether this arrangement meets that standard is not addressed by Google's opt-out mechanism.