Who shapes AI, and on whose terms

The United Nations has appointed MTN Group chief executive Ralph Mupita to its AI for Good Global Commission, a 44-member body tasked with shaping how artificial intelligence is developed and governed in ways that are fair and broadly beneficial, according to Business Day. Mupita joins a panel that includes senior international figures such as Unesco director-general Khaled El-Enany. The appointment places a prominent South African business leader at the table where global AI governance norms are being shaped, giving South Africa a direct voice in those discussions.

TechCentral reports that China's government has published a five-year AI plan that includes a deliberate push to export AI technology and humanoid robots (machines built to move and work like people) to Global South countries, a group that includes South Africa. The plan also targets a doubling of clean energy capacity to power the computing infrastructure that AI systems require. For South Africa, the report suggests the country is an intended market for Chinese AI products and infrastructure, raising questions about technology dependence, standards, and the terms on which that technology would arrive.

South African fintech lender Bridgement has secured R330 million from local banks to expand its AI-driven small-business loan platform, according to ITWeb. The company uses artificial intelligence (software that analyses data to make decisions) to assess loan applications, a model that could widen credit access for small businesses that traditional bank scoring often excludes. The scale of the funding round signals growing appetite among established South African banks for AI-based lending infrastructure.

Pick n Pay's on-demand delivery service, asap!, has launched what Business Day reports is the first fully integrated AI-powered shopping assistant on a grocery delivery platform in South Africa, allowing shoppers to get personalised help while browsing and ordering. The move places one of the country's largest retailers among a growing number of businesses using AI tools (software that can understand and respond to natural language) directly in consumer-facing services. For South African shoppers, it marks an early example of this kind of technology shaping everyday retail interactions, though Business Day's report offers limited detail on how the system works or what data it draws on.

The UK's Financial Conduct Authority has warned that regulators are in an "arms race" to keep pace with the rapid spread of AI in financial services, with millions of people already using tools such as ChatGPT and Claude (large language models, meaning AI systems trained on vast text to generate responses) for personal finance decisions, according to Ars Technica. A senior FCA official, writing a commissioned report on the issue, called for greater regulatory powers and urged authorities to consider whether existing rules should apply to these tools. South Africa's own financial regulators face a comparable gap, as AI-driven finance tools reach consumers here too, and the UK's approach is likely to shape the frameworks that SA institutions and firms with UK exposure will need to follow.

TechCentral argues that one of AI's most consequential effects in South Africa may not be job losses among established workers, but the disappearance of entry-level roles before young people can get a foothold in the labour market. With youth unemployment already at crisis levels, the concern is that AI tools are absorbing the routine, lower-skilled tasks that have historically served as the first rung of a career. The piece frames this as a structural risk to economic mobility that deserves attention beyond the usual debate about automation displacing existing jobs.

Writing in Business Day, Garon Campbell argues that AI tools are shifting what employers look for in creative professionals: the advantage is moving toward people who combine specialist skills with curiosity, a willingness to collaborate across disciplines, and comfort working alongside AI systems. For South Africa's creative sector, where employment in advertising, design, and media is already under pressure, the piece suggests that adaptability and breadth of thinking may matter as much as technical craft in hiring decisions going forward.

New environmental disclosures from Google and Amazon show that the rapid expansion of AI infrastructure is pushing up electricity use, water consumption and carbon emissions faster than either company's clean-energy commitments can offset, according to Memeburn. Google's electricity demand rose 37% last year and its greenhouse gas emissions hit their largest annual increase on record, while Amazon added more than 1.2 gigawatts of data centre capacity in the final quarter of 2025 alone. For South African businesses and developers, Memeburn notes two direct angles: Amazon Web Services already runs a Cape Town region serving local companies, and if data centre operating costs keep climbing, that pressure may eventually reach customers through higher prices or usage limits – a concern with added weight in a country where electricity supply remains an economic variable.

OpenAI is reportedly in early discussions about offering the US government a 5% equity stake in the company, according to a Financial Times report cited by Memeburn. The proposal, which has not been finalised, would create a direct financial link between one of the world's most influential AI developers and the US state, raising questions about whether regulators could oversee the company independently. For South African businesses, universities, and developers who rely on US-built AI tools, Memeburn notes that closer ties between Washington and major AI labs could affect access to advanced models, pricing, and the terms under which those tools are available outside the United States.

Anthropic, the company behind the Claude AI assistant, quietly removed hidden tracking code from its Claude Code product after a security researcher discovered it had been secretly monitoring users in China since March, according to Ars Technica. The code used a technique called prompt steganography (embedding hidden markers inside text that most users would not notice) to collect information about users' time zones, internet connections, and possible links to Chinese AI companies. Anthropic confirmed the tracker was added as an experiment to curb account abuse and protect against unauthorised use of its models, but the episode raises questions for South African organisations and individuals using AI tools about what data major AI providers may collect without clear disclosure.