Whose AI is it, and does it work for us

Writing in Business Day, legal scholar Letlhokwa Mpedi and brand strategist Thebe Ikalafeng argue that Africa has no adequate laws to stop AI systems from exploiting the continent's cultural heritage and intellectual property. They contend that AI models trained predominantly on Western data misrepresent African cultures and enable appropriation without consent or compensation. For South Africa, the piece points to a concrete regulatory gap: as AI-generated content draws on local languages, music, art, and traditional knowledge, there is currently no legal framework that gives communities or rights-holders meaningful recourse.

African heads of state, including South Africa, are shifting their approach to artificial intelligence (AI) from ethics and startup support towards hard infrastructure: data centres, computing capacity, sovereign data systems, and locally built language models (AI systems trained on African languages and data). TechCabal reports that 30 leaders signed the Africa Forward Declaration in Nairobi in May, committing to mobilise public and private investment across the full digital and AI stack, while acknowledging that the continent hosts less than 1% of global data-centre capacity despite rapid data growth. For South Africa, which is advancing its own national AI policy, the declaration frames AI infrastructure as a strategic and economic priority on par with roads and ports, and signals that the country's choices about data ownership, compute investment, and foreign partnerships will shape its position in an increasingly contested global technology landscape.

The White House has directed US intelligence and military agencies to accelerate their use of artificial intelligence (AI) in defence, cybersecurity and intelligence work, while stating the technology must not enable unlawful surveillance or suppress free speech, Reuters reported. The policy also asks leading AI developers to voluntarily submit their most capable models for government security testing before public release, and gives the Defence Secretary 90 days to update rules on weapons systems that can act without direct human control. Memeburn notes that as US norms around defence-grade AI shape what the largest technology companies build, those tools can later appear in commercial products used by South African banks, telecoms and public agencies, raising questions about privacy, accountability and who sets the rules for countries that adopt rather than develop these systems.

MTN Group has set a target of R30-billion in value from artificial intelligence by 2030, according to TechCentral, with the pan-African telecoms company focusing its initial efforts on using AI internally to improve its own operations before turning to customer-facing applications. The company has ranked its AI bets by how confident it is in executing each one, signalling a staged rather than all-at-once approach. For South Africa, where MTN is headquartered and operates at scale, the strategy offers an early indication of how one of the continent's largest telecoms groups plans to deploy AI across its networks and business.

Global stock markets have surged to record highs on the back of intense investor enthusiasm for artificial intelligence, but TechCentral reports that analysts are raising concerns about speculative excess and whether electricity supply can keep pace with the computing demands AI requires. For South African investors and energy planners, both risks are relevant: local portfolios with exposure to global tech stocks face the same correction risk, and South Africa's already strained power grid sits in the background of any ambition to expand AI infrastructure here.

Researchers and local technology vendors warn that AI voice agents (software that handles spoken interactions, such as call-centre bots) built for global markets routinely misread South African accents, struggle with the country's eleven official languages, and are slowed by local network latency, according to TechCentral. The result is that many South Africans find these systems unreliable or unusable in practice. The report points to a gap between where the technology performs well and where it is being deployed here.

South African researchers from Stellenbosch University, the University of Fort Hare, and the University of Cape Town have identified artificial intelligence as both an opportunity and a risk in biodiversity conservation, according to a horizon scan published by Stuff South Africa. AI tools can help process large volumes of ecological data, track animal movements through camera recognition, and detect illegal wildlife trade online, but the researchers warn that these systems can generate false information, reflect biases toward data from wealthier institutions, and miss context that a skilled ecologist working on the ground would catch. The team calls for clear regulation, validation standards, and mandatory disclosure of how AI tools are trained, arguing that uncritical use risks sidelining indigenous knowledge and local communities in conservation decisions across Africa.

Writing in Business Day, Rufaro Mafinyani explains how machine learning systems (software that finds patterns in large amounts of data and uses them to make predictions or decisions) can go wrong when the data they are trained on is biased or out of date. The piece argues that these hidden flaws can produce unfair or inaccurate outcomes, with real consequences for the people and institutions that rely on such systems. For South African readers, the concern is practical: as more local organisations adopt these tools, understanding where they can fail is a necessary starting point for using them responsibly.

Google has expanded its Gemini AI assistant in the Chrome browser to South Africa, as part of a broader rollout across Africa, the Middle East, and Latin America, according to Hypertext. The feature, built into the browser's side panel, lets users summarise web pages, compare information across open tabs, and interact with Google services such as Gmail, Maps, and Calendar without leaving the current page. South African Chrome users can now check whether the feature is available in their region, though how widely it will be adopted remains to be seen.