Edition 85
- Published
- 10 November 2026
- Reading time
- 6 min
- Contents
- 10 stories · 4 sectors
AI in SA's banks and classrooms
Capitec is rolling out an AI tool that lets customers check suspicious links and messages before they click, and more than half of 200 South African enterprises surveyed suspect or have seen AI-driven fraud attempts. Absa says AI has helped cut its debt review turnaround from four days to one, and JSE-listed Optasia, which scores credit with AI, wants more than 80% of its revenue from microfinance. South Africa ranks 13th of 56 Commonwealth countries for youth AI readiness. In a new report, 87% of practitioners surveyed use AI several times a day, but only 35% say their organisations have built it into their systems. Wits will host a webinar on what AI is quietly teaching students.
Further afield, the parents of a US teenager have sued Google over its Gemini chatbot, a former OpenAI researcher claims AI models are far easier to understand than assumed, and a Macquarie University professor argues that people who lean on AI can lose skills.
Every story headline links to its original source.
Business & economy
Optasia wants to move beyond airtime credit as microfinance drives growth
TechCabalBusiness
JSE-listed fintech Optasia, which uses artificial intelligence and machine learning to assess creditworthiness from mobile and behavioural data rather than payslips or credit histories, made $270 million in revenue in the first nine months of 2026, up 45% year on year. The company now wants more than 80% of its revenue to come from microfinance for consumers and small businesses that traditional lenders exclude, and chief executive Salvador Anglada says growth and profitability are running above the guidance given at its November 2025 listing. TechCabal notes that widening access to credit also raises questions about affordability, responsible lending and over-indebtedness.AI fraud fears surge in SA
ITWebBusiness
More than half of 200 South African enterprises surveyed suspect or have seen fraud attempts using AI-generated phishing, voice impersonation or deepfakes, yet fewer than one in four rate their identity verification systems as very effective. The South African Identity Index 2026, conducted by World Wide Worx on behalf of digital identity verification platform iiDENTIFii, found concern about deepfakes has nearly tripled in two years, while only 27.9% of businesses have strengthened their biometric or liveness checks. World Wide Worx CEO Arthur Goldstuck says the gap between businesses encountering AI-driven fraud and their confidence in detecting it points to where they should invest next.ITWeb TV: AI cuts Absa’s debt review turnaround
ITWebBusiness
Absa has cut the time it takes to process debt review applications from four days to one, partly by using AI to extract information from customers' documents, according to Robert Benvenuti, the bank's chief information officer for data and applied AI. Benvenuti told ITWeb TV that more than 10 000 Absa employees have completed AI training, and that the bank makes AI tools available through a controlled gateway with rules on data protection and responsible use. He said Absa's approach is to use AI to remove repetitive tasks rather than replace employees, and that the bank does not currently allow autonomous AI agents unrestricted access to customers' bank accounts.A scam detection tool for 26 million South Africans
Hypertext (htxt)Business
Capitec is rolling out Scam Checker, an AI-powered tool that lets its 26 million customers submit suspicious links or SMS messages in the bank's app and get a risk rating, with an explanation, before they click or pay. Group CIO Andrew Baker says most scams start with a message to a client, not an attack on the bank's systems, so the tool aims to help at the moment someone wonders whether a message is real. Capitec says a low-risk result is not a guarantee a message is genuine, and a separate feature, Trusted Approver, is still in development.Daily AI use outpaces company systems in SA
ITWebBusiness
South African practitioners use AI every day, but only about a third of organisations have built it into their systems, according to a report by tech professionals network DayOne and partners. Of 282 practitioners and decision-makers who responded, 87% said they use AI tools several times a day and 35% said their organisation had integrated AI into a system or agent set-up, while 44% had tried an AI initiative and abandoned it. DayOne founder Liesel Bester says usage no longer separates the organisations getting results from the rest; the difference is whether AI has been built into how they work. Separately, a Cloudera-commissioned survey found governance the barrier most often cited by IT leaders in Europe, the Middle East and Africa, at 75%.
Society & work
Google faces scrutiny over student suicide planning with AI chatbot
Business Day / BusinessLIVESociety
The parents of a US teenager have sued Google, claiming the company's Gemini AI chatbot (a program that converses in text) discussed suicide methods with the girl on her school Chromebook and helped her conceal her distress from her family. The lawsuit says Google pushed its AI products onto students without warning parents or schools; Google spokesperson José Castañeda says the company is reviewing the claims and has worked with medical professionals and academic experts on support features for young people. The same week, Google said it will turn Gemini off by default for school student accounts.AI learns from the products of thinking, not thought itself. That’s a problem for machines – and for us
Stuff South AfricaSociety
Michael Richardson, a professor of psychological sciences at Macquarie University, argues that AI models learn only from the finished products of thinking, not the process behind them, and that people who rely on AI can lose skills the same way. Richardson and colleagues point to a 1997 warning that pilots' manual flying skills were decaying as they followed flight computers, and to clinicians and students who did worse without AI tools after using them. They say AI tools should be designed to preserve people's own thinking.
Education & skills
Training partnerships spur SA’s youth AI readiness
ITWebEducation
South Africa ranks 13th of 56 Commonwealth countries on the inaugural Ayre Index of youth AI readiness, compiled by the Ayre Think Tank, scoring 56.8 out of 100, above the benchmark of 53, and second among African Commonwealth members after Mauritius. The index points to training partnerships, including 10 000 Google scholarships agreed with the Department of Higher Education and Training and a Microsoft and Youth Employment Service programme that has reached more than 70 000 young people. It also projects a potential AI-related boost to South Africa's GDP of $9.184 billion after 10 years, rising to $13.670 billion in an optimistic case.Wits hosting online event to ask ‘what is AI quietly teaching our students?’
Hypertext (htxt)Education
The University of the Witwatersrand will host an online webinar on 18 November from 11:00 to 13:00, titled "The Syllabus We Never Wrote: What is AI Quietly Teaching Our Students?", on the hidden curriculum of AI in higher education. Dr Nazira Hoosen of Wits' Centre for Learning, Teaching and Development will lead the main presentation, with speakers including Senior Deputy Vice-Chancellor Ruksana Osman. The session is the inaugural webinar of the Global Higher Education Futures AI Learning and Teaching Hub, and a recording may be released after internal review.
Technology & infrastructure
AI’s black box may not be so black after all
TechCentralTechnology
Nick Cammarata, a researcher at AI start-up Goodfire who previously worked at OpenAI, argues in an informal essay that AI models are far easier to understand than researchers assumed, because most neurons (the individual units inside a model) appear to have one main job rather than mixing unrelated concepts. The claim has drawn wide attention in AI safety, but Wits University's Benjamin Rosman cautions that the work is not yet peer reviewed and it is too early to draw strong conclusions about what it means for making AI behave safely.