Edition 72

Published
6 October 2026
Reading time
6 min
Contents
11 stories · 4 sectors

AI starts deciding, and South Africa drafts the rules

This edition looks at AI moving from talk to decisions that affect people directly. Home Affairs plans to let machine learning make first calls on visa applications, while the Department of Communications & Digital Technologies rewrites its AI white paper after the first draft was found to be AI-generated with invented citations. An investment chief weighs which South African jobs are most exposed, and tourism bodies grapple with how AI trip planners represent the country. Beyond our borders, the US and China explore an AI crisis hot line, a leaked Anthropic filing raises questions about who controls frontier AI, and reports of a chatbot error that nearly triggered a military incident show what happens when machine guesses are mistaken for vetted intelligence.

Every story headline links to its original source.

Policy & governance

  1. Balancing the AI revolution: Lessons from BRICS New Delhi Summit

    ITWebPolicy

    The 18th BRICS Summit in New Delhi ended with a declaration calling for wider access to AI resources, stronger safeguards, and more inclusive participation in how the technology is developed and governed. Sandile Nene, acting deputy director-general of the Government Communication and Information System, writes that without safeguards AI could deepen inequalities and leave the Global South as users of technology built elsewhere. President Cyril Ramaphosa proposed that BRICS set up an international mechanism for the independent scientific evaluation of AI.
  2. AI policy gets a second chance

    Business Day / BusinessLIVEPolicy

    The department of communications & digital technologies will produce a new draft of its AI white paper, after the first version was withdrawn when it was found to be AI-generated with hallucinated citations (invented references). Deputy minister Mondli Gungubele said the final draft is due for ministerial input before the end of November and should go to Cabinet by the end of the financial year. He said AI is already operating in South Africa with or without a policy, and that the policy is there to set the limits it operates within.
  3. Machine learning gets first say on who gets into SA

    ITWebPolicy

    Home Affairs plans to use machine learning, a form of artificial intelligence that learns from data, to make yes-or-no decisions on visa applications through its Electronic Travel Authorisation (ETA) system, with a human official stepping in only when the system flags an anomaly. Minister Dr Leon Schreiber says the system, now live at four airports and covering short-stay visas from 53 countries, will expand in early 2027 to study and work visas, and he argues it replaces opportunities for corruption and delay with automation. He says approved authorisations are issued within 24 hours, down from processes that previously took weeks.
  4. US and China explore AI ‘hot line’ amid safety concerns

    IOL (Independent Online)Policy

    The United States has proposed a direct communication channel with China on artificial intelligence, to notify each other of AI crises or misuse, US Treasury Secretary Scott Bessent said after talks with Chinese officials ahead of a Trump–Xi summit. One analyst likens it to the Cold War telephone "hot line" set up after the Cuban missile crisis, though the consultancy Teneo says neither government is yet willing to commit to binding limits on AI development.
  5. A Chatbot Almost Started a War With China. The Scariest Part Wasn’t the AI’s Mistake

    MemeburnPolicy

    According to reports, a US military analyst fed a chatbot a mix of public and classified information about a Chinese ship, and the chatbot wrongly concluded it was carrying nuclear weapons parts bound for Iran; US aircraft were already airborne to intercept the vessel before officials found the report was false. Memeburn argues the deeper failure was a later step, in which the chatbot's guess was reformatted into a standard intelligence report that looked like any human-vetted assessment, so no reader could tell its core claim came from a machine. Three US senators, including Mark Warner, have asked for an investigation into how AI errors could prompt military action against a nuclear-armed country. The Pentagon has not publicly explained the incident.

Business & economy

  1. SA Tourism faces AI challenge

    Business Day / BusinessLIVEBusiness

    South Africa's digital readiness for tourism fell from 5.19 to 5.12 out of seven in the World Economic Forum's 2026 travel and tourism index, dropping the country from 56th to 65th place. Arthur Goldstuck argues that small tourism businesses here struggle to reach travellers because their information is poorly represented online, and points to Germany's shared tourism database, which lets AI trip planners draw on more than 500,000 records, as a model South Africa could follow. South African Tourism has begun with Siyanda, an AI assistant offering US and Canadian travellers personalised local itineraries, but its chief strategy officer Siya Mthethwa has warned that poorly represented local details give AI room to make things up.
  2. SA’s big earners in the crosshairs as AI adoption accelerates

    Business Day / BusinessLIVEBusiness

    Ashburton Investments chief investment officer Patrice Rassou argues that AI is most likely to affect South Africa's financial, business and professional services sectors, where highly paid knowledge workers earn a large share of the national wage bill. Rassou cites a Goldman Sachs index that places South Africa in the middle of the pack among emerging-market laggards on AI adoption, and an Anthropic scenario in which the US economy grows while white-collar jobs fall 4% by 2030. Rassou expects sectors such as agriculture, mining, construction and hospitality to be less exposed to direct substitution by generative AI.
  3. Anthropic Told Investors Its AI Could End Humanity. The Fine Print That Matters More Is Who Holds the Off Switch

    MemeburnBusiness

    A leaked draft of the IPO filing of Anthropic, the company behind the Claude chatbot, warns investors that advanced AI models could show self-preserving behaviour, resist shutdown and pose catastrophic risks to humanity, ahead of a listing that could value the company at more than $2 trillion. The filing also sets out an unusual control structure: the seven co-founders and a Long-Term Benefit Trust would keep most of the voting power, so public shareholders would own a share of profits but could not vote out the board. The company has signed $518 billion in computing commitments, about 80% of which it cannot cancel, which the piece frames as the tension between its safety mission and the cost of staying at the frontier.
  4. AI spending now dwarfs the railway and dot-com booms

    TechCentralBusiness

    Cumulative global spending on data centres alone (the large facilities that house and run AI) could top US$30-trillion by 2050, a PwC projection says, which PwC says dwarfs what was spent in the railway and dot-com booms, even after inflation. Economists at JPMorgan and Bain & Company say the productivity gains needed to justify this outlay have not yet appeared, with Bain estimating AI firms must find more than $4.2-trillion in new revenue over five years to fund the build-out.

    Also reported by News24 / Fin24

Education & skills

  1. How AI is saving Nigerian lecturers days of preparation

    IOL (Independent Online)Education

    A lecturer at a Nigerian finance academy says an AI platform that generates course slides, videos, assignments and exams has cut preparation that used to take days down to minutes. Idris Abba, a lecturer at Abuja's Santuval Academy, says the platform's AI assistant also marks coursework and gives students scores and feedback. Umar Adam Ibrahim, a lecturer at Nile University, says such tools can free lecturers' time for research and mentorship, and give students learning paced to their own speed.

Technology & infrastructure

  1. Land availability gives SA edge in data centre race

    ITWebTechnology

    Research firm DC Byte has identified South Africa as a next-wave market for data centre growth in the Europe, Middle East and Africa region, saying its available land could attract the larger facilities that AI workloads demand. Lead analyst Christian Tshishiku said power supply and delivery timelines remain the deciding factors, noting that Africa holds just 3% of the region's data centre capacity. He said African markets can attract new investment only if they meet the infrastructure requirements driving investors to look beyond established European hubs.

Get the next edition

Confirm by email Published when there’s real news Unsubscribe anytime