AI in the workplace, the wallet, and the war room

Anthropic, the US company behind the Claude family of AI models, took its most capable new models offline last week after a United States government export-control order barred foreign nationals from using them, according to Ars Technica. The affected models include one Anthropic has described as having advanced ability to find and exploit software vulnerabilities, a capability the company itself flagged as double-edged: useful to security researchers, but potentially dangerous in the wrong hands. South African researchers, developers, and institutions that rely on access to frontier AI models (the most powerful systems at the leading edge of the field) would be directly affected by a blanket foreign-national restriction of this kind.

Investec, the dual-listed South African bank, is rolling out Microsoft's AI assistant tool to every employee across its South Africa and UK operations, according to TechCentral. The move makes Investec one of the larger local financial institutions to deploy AI at this scale in the workplace, putting the technology directly in front of staff rather than limiting it to specialist teams. How the rollout affects day-to-day work, and whether employee concerns are part of the process, is not detailed in the report.

Research firm Gartner predicts that by 2028 the cost of AI-assisted coding tools (software that helps developers write and review code automatically) will approach what companies pay developers in salaries, according to TechCentral. The publication suggests this crossover could arrive sooner in South Africa, where developer pay scales differ from global averages, raising questions about the economics of hiring versus licensing AI tools for local technology teams.

A PwC Global report ranks mining last among major industries in readiness to adopt AI, according to Business Day. For South Africa, where mining remains a central part of the economy and a major employer, the finding points to a concrete gap: local mining companies may be slower than peers in other sectors to use AI tools for tasks such as safety monitoring, equipment maintenance, and ore processing. The report's implications extend to workers and policymakers as well as company boardrooms.

Alex Thomson, co-founder of South African insurance technology company Naked Insurance, spoke on a Business Day podcast about how the company is using ChatGPT (a general-purpose AI tool made by OpenAI) in its business. Naked operates in a consumer-facing sector where AI-driven decisions can directly affect policyholders, making its approach of interest to South Africans watching how local financial services companies adopt this technology.

South African travel platform FindMy is using AI agents (software that can hold a conversation, make decisions, and complete tasks on a user's behalf) to handle holiday planning end-to-end, replacing the familiar process of opening multiple browser tabs and comparing results manually, according to IT News Africa. Rather than returning a list of links, the system takes a traveller's requirements through a conversation and produces a plan and booking directly. The story illustrates how conversational AI is beginning to reach South African consumer services, with the December holiday season cited as an early test of whether local users will adopt the approach at scale.

A Cape Town couple was defrauded by scammers who used AI-generated voice and video cloning to impersonate their son, showing him apparently hospitalised and in urgent need of money, according to Moneyweb. Thalia Pillay of fraud advisory firm Orca Fraud describes the case as part of a growing pattern of AI-enabled scams in which the technology is used to fabricate convincing likenesses of people known to the victim. The incident illustrates a concrete risk for South African consumers as these tools become easier to access and deploy.

Writing in Business Day, commentator Rufaro Mafinyani examines how AI systems are increasingly being used to read and analyse workplace communications, documents, and decisions inside organisations. The piece raises questions about what this means for employees and institutions, covering the governance and privacy implications of AI that monitors or processes internal text at work. For South African organisations, the argument points to a need to understand what these tools are doing with staff communications before deploying them.

The Institute of Information Technology Professionals South Africa (IITPSA), the country's professional body for ICT practitioners, is conducting a skills survey to measure how AI is affecting jobs in South Africa's technology sector, according to ITWeb. The survey aims to build a clearer picture of which roles are changing, which skills are in demand, and where gaps may be opening up. For workers and employers in the sector, the findings could inform training decisions and workforce planning at a time when AI is reshaping many technical roles.

Anthropic, the US company behind the Claude AI model, has accused Chinese technology group Alibaba of running a large-scale, unauthorised operation to extract Claude's capabilities through nearly 25,000 fraudulent accounts, generating 28.8 million interactions between April and June, according to a letter Anthropic sent to US senators and White House officials and reported by Bloomberg. The technique involved is known as distillation, where a developer trains a new AI system on the outputs of an existing one to replicate its capabilities at far lower cost, without permission or the original research investment. For South Africa, the story signals growing legal and regulatory pressure around how AI models may be used or copied, a dynamic that could affect which AI tools remain available here, on what terms, and at what cost.