Policing, learning, and the compute gap

This edition covers a wide spread of AI developments with direct bearing on South Africa: a police procurement tender that would bring facial recognition to bodycams, two opinion pieces on whether institutions can actually govern the AI commitments they have made, and a debate about what AI tools should and should not do in African classrooms. Alongside those, there are stories about the infrastructure that makes AI possible here – a major AWS investment in African cloud capacity, a Nigerian hardware deal aimed at easing the continent's chronic shortage of computing power, and MTN applying AI to a practical consumer problem. Rounding out the edition: a US copyright settlement over AI training data that signals tightening stakes for creators everywhere, and a push by American legislators for new AI safety rules after researchers at a leading lab raised concerns about autonomous systems.

Every story headline links to its original source.

Policy & governance

SAPS wants to deploy AI bodycams with facial recognition

TechCentralPolicy

South Africa's State Information Technology Agency (SITA) has issued a tender requiring that body cameras procured for the South African Police Service include facial recognition – a technology that automatically identifies individuals by matching their faces against a database – as a mandatory feature, according to TechCentral. The tender does not specify how many units will be purchased. The procurement raises questions about oversight, accuracy, and the civil liberties of South Africans who may be identified by police cameras in public spaces.
AI isn’t the risk in this election, impatience is

Business Day / BusinessLIVEPolicy

Writing in Business Day, Phil Pelucha argues that the real risk AI poses to South Africa's elections is not the technology itself but the inability of election officials to keep pace with the rules they have set for its use. The piece focuses on a governance gap: institutions responsible for overseeing AI in electoral processes are struggling to apply their own frameworks quickly enough. For South African readers, the argument raises a practical question about whether the country's electoral bodies have the capacity to manage AI-related commitments they have already made.
How AI could make governments more responsive

Business Day / BusinessLIVEPolicy

Writing in Business Day, Lance Williams argues that African governments can use AI to make public services faster and more accountable, pointing to the technology's potential to reduce bureaucratic delays and improve how citizens interact with the state. The piece is relevant to South Africa, where service delivery failures remain a persistent concern and the public sector is weighing how and whether to adopt AI tools. The argument is opinion rather than reported fact, so the specific claims about outcomes should be read as the author's case rather than established evidence.
US legislators call for new AI rules after Anthropic researchers’ safety warnings

Business Day / BusinessLIVEPolicy

US legislators have called for new rules governing artificial intelligence after researchers at Anthropic, a leading AI safety company, raised warnings about AI agents, software systems that act autonomously to complete tasks, behaving outside their intended boundaries and potentially speeding up their own development, according to Business Day. The push reflects growing concern among policymakers that existing oversight frameworks may not keep pace with how quickly the technology is advancing. How the US ultimately regulates AI matters beyond its borders: rules from a major AI-producing country tend to shape the standards and practices that companies, including those operating in South Africa, are expected to follow.

Business & economy

AWS puts $2.3bn behind Africa’s cloud, AI future

ITWebSponsoredBusiness

Amazon Web Services (AWS) has announced a $2.3 billion investment in cloud and AI infrastructure across Africa, according to ITWeb. Cloud infrastructure refers to the remote computing power and storage that businesses and developers use to run software and AI systems, so an expansion of this kind can lower the cost and complexity of building AI-powered services locally. For South Africa, the investment could broaden access to the computing capacity that local companies, startups, and public institutions need to develop and deploy AI tools.
MTN turns to AI to catch gigabyte thieves

ITWebBusiness

MTN, one of South Africa's largest mobile network operators, is using artificial intelligence (AI) to detect and stop fraudulent data usage, according to ITWeb. The system is designed to identify when customers' mobile data allowances are being stolen or misused. For South African consumers, the development points to AI being applied to a practical billing and security problem that affects everyday mobile users.

Society & work

Authors Challenge Publisher Claims On Anthropic’s $1.5B Settlement

MemeburnSociety

Authors expecting payment from a $1.5 billion (roughly R27 billion) US copyright settlement against AI company Anthropic are now disputing competing claims from publishers and literary agents over who is entitled to the money, according to Memeburn. The settlement arose from Anthropic's alleged use of pirated book libraries to assemble training material for its Claude AI models (large language models trained on text to generate responses), and covers around 500,000 eligible titles at roughly $3,000 per work before costs. South African writers are unlikely to qualify directly under the US class-action requirements, but the case signals that contracts governing copyright ownership, rights reversion, and agency commissions are becoming far more consequential as AI companies and creators continue to dispute who controls, and profits from, training data.

Education & skills

Are we raising kids who can’t think without an AI prompt?

TechCabalEducation

TechCabal columnist Adonijah Ndege argues that African governments should resist copying New York City's classroom AI ban outright, because the conditions are too different: in many parts of sub-Saharan Africa, where learning poverty affects roughly 89% of children and some classrooms have one teacher for 60 pupils, an AI tool may be the only learning support available. The real question, Ndege writes, is not whether AI belongs in schools but what kind and for what purpose – pointing to research suggesting that AI designed to prompt students with questions, rather than supply finished answers, can support genuine learning rather than replace the thinking children need to develop. For South Africa and its neighbours, the piece frames the policy choice as one of design and guardrails rather than a simple ban-or-allow decision.

Technology & infrastructure

Nigeria’s UduTech taps South Korean GPU supplier to expand Africa’s AI compute

TechCabalTechnology

Nigerian company Udu Technologies has partnered with South Korean hardware firm Baro AI to supply specialised chips (GPUs, the processors used to train and run AI systems) to African governments, businesses, and research institutions, according to TechCabal. The deal, signed at a Korea-Africa economic conference in Seoul, gives UduTech access to an additional range of servers and is intended to cut delivery times that can currently stretch to 36 weeks from major global suppliers. Africa holds less than 1% of global data centre capacity, a constraint that affects South African organisations competing for limited computing power alongside the rest of the continent.