Global AI rivalries and what they mean here

South Africa's Minister of Science, Technology and Innovation, Blade Nzimande, met with his Chinese counterpart at the 2026 World Artificial Intelligence Conference in Shanghai to explore deeper collaboration in AI, research, and related fields, according to Hypertext. China has also invited South Africa to become a founding member of the World Artificial Intelligence Cooperation Organisation (WAICO), a new body aimed at international AI cooperation. The development signals a potential shift in where South Africa sources AI systems and governance frameworks, with most AI services currently coming from Western providers and Huawei among the few Chinese firms already active in the local market.

The International Monetary Fund has warned that Africa risks being left out of the economic gains from artificial intelligence unless the continent significantly improves its electricity supply, internet access, and digital skills, according to Business Day. The report frames these infrastructure gaps as the primary barriers to AI adoption across African economies. For South Africa, which faces its own well-documented constraints in power supply and skills development, the findings point to concrete preconditions that would need to be met before AI can deliver broad economic benefit.

Chinese regulators are in talks with major AI companies including Alibaba, ByteDance and Zhipu about restricting exports of the country's most advanced AI models to Western markets, according to TechCentral. The move would limit who can access frontier Chinese AI systems, the large-scale software trained on vast data that powers tools like chatbots and image generators. For South African organisations that use or are considering Chinese AI products, such restrictions could narrow their options, though the practical effect here is not yet clear.

Microsoft is training its sales teams to position the company's own AI products as direct competitors to OpenAI and Anthropic, even though both companies supply models that Microsoft also resells through its platforms, according to reporting by Memeburn citing TechCrunch and Bloomberg. The pitch centres on lower costs, tighter security controls, and selling a complete business platform rather than individual AI tools, with Microsoft aiming to own the layer around its partners' models, including the user interface, company data, and billing. For South African organisations, Memeburn notes this rivalry should prompt harder procurement questions: businesses should test rand-based costs, local language performance, and how easily data can move between providers before committing to any single vendor's platform.

TechCentral, in content supplied by Domains.co.za, explains how data poisoning works: when the information used to train an AI system is deliberately corrupted, the system can produce biased, inaccurate, or manipulated outputs without the business using it being aware. The piece is aimed at South African businesses that rely on AI tools and want to understand this category of security risk. Readers should note the content originates from a commercial domain and hosting provider rather than an independent editorial source.

Elon Musk's AI company xAI has sued one of its own users in a Texas federal court, alleging that the man used Grok, xAI's AI chatbot, to create and distribute child sexual abuse material by repeatedly changing his requests after the system refused them, according to Memeburn citing a Reuters report. The case centres on a contract-law argument: xAI contends the user broke its terms of service, placing responsibility on him rather than on the platform's design, though xAI separately faces lawsuits claiming Grok's own image tools made such abuse too easy to carry out. For South Africans, Memeburn notes that local law already criminalises this material and that the Film and Publication Board operates a reporting hotline, but the case raises a broader question about whether global AI tools do enough to prevent harm before it occurs, rather than responding after the fact.

South African director Neill Blomkamp, known for District 9, has released a 13-minute science-fiction short called Nightborne made entirely using ByteDance's Seedance 2.0, a tool that generates video footage from text descriptions. Blomkamp described the project as a test of what generative AI (software that produces new content from written prompts) can do, and said he wants to use the technology for a full-length feature film. The Verge reports that the result has drawn criticism for lacking the substance of conventional filmmaking, raising questions about the creative and commercial role of AI-generated video in the film industry.

Amazon Web Services (AWS) has launched a programme in Cape Town in which more than 30 schools are taking part, with learners aged 13 to 16 using a generative AI design tool (software that produces images or concepts from text prompts) to create sneaker designs using recycled materials, as part of a global competition run with footwear brand Brooks Running, Hypertext reports. The Cape Town facility is the first of AWS's dedicated STEM learning spaces on the African continent, and the company says it aims to reach 10 000 South African learners across six provinces through partnerships with local non-profit organisations, though no firm timeline for that target has been given. The initiative is one signal of how large technology companies are positioning AI as a practical classroom tool for young South Africans, though the programme is AWS-led and the reported benefits come from the company itself.