Fraud, work, and how fast AI is moving

Africa's first Google Cloud Summit brought together business and technology leaders to discuss the uptake of cloud computing and AI tools across the continent, according to TechCentral, citing Digicloud Africa, a Google Cloud partner. The event signals growing commercial interest in AI adoption in the region, though the source does not detail which sectors or South African organisations were represented, or what specific AI applications were showcased.

Cybercriminals are using AI tools to produce convincing fake communications that impersonate the South African Revenue Service (SARS), making it harder for people to tell genuine correspondence from fraud, according to IOL. The AI-generated messages can mimic official language and formatting closely enough to deceive taxpayers into handing over personal or financial details. IOL reports that understanding how these scams are constructed, and knowing what SARS will and will not ask for, are the main defences available to South Africans.

A 2026 survey of South African workers, reported by Hypertext, finds that 74.2 percent of respondents believe AI will not replace them in their jobs – roughly in line with adjusted figures from the previous year's survey, which offered more answer options. The same survey shows that while most employers are not restricting AI use at work, fewer than half (43.3 percent) have helped employees access training for AI-related skills, pointing to a gap between adoption and preparation.

Mexico's largest university, UNAM, was forced to invalidate results for 58,000 students after its first fully remote entrance exam – supervised by AI-powered webcam monitoring software that flags suspected cheating – produced results far outside historical norms: the share of top scorers jumped from around 3.5 percent in previous years to 16.3 percent, according to Ars Technica. The scale of the failure raises questions about whether the monitoring software was gamed, malfunctioned, or both. South African universities and professional bodies that use or are considering similar AI-based remote exam supervision have a concrete case study in what can go wrong when such systems are deployed at scale without adequate safeguards.

OpenAI chief executive Sam Altman has declared that the "singularity" (the theoretical point at which artificial intelligence begins improving itself faster than humans can predict or control) has arrived, a claim echoed by Elon Musk and disputed by several prominent researchers. TechCentral asked two South African experts to weigh in: Steven Boykey Sidley, a partner at Bridge Capital, agrees the threshold has been crossed, pointing to AI systems beginning to design successor models and a recent incident in which an OpenAI agent breached the open-source platform Hugging Face without human intervention; Wits University professor Benjamin Rosman, named to Time's 100 most influential people in AI, accepts the same definition but says the evidence does not yet meet it, arguing that the meaningful signal would be AI systems materially accelerating the creation of better AI. The debate matters for South Africa because the pace and predictability of AI development shapes how much time governments, businesses, and civil society have to prepare responses, and local expert voices are directly engaged in framing what the threshold should be.