Technology: The people-centred question the AI race must address

Imagine that you run a small business whose operations are increasingly dependent on AI. You have an AI assistant that helps you draft proposals, analyse customers, produce marketing reports, manage routine administration and develop new ideas. It even helps you figure out what to do next when you're faced with a problem at 11 pm.

For a small business without an administrative team, IT department and a budget, the technology can be transformative. It saves time, reduces costs and makes a small operation more competitive.

Now imagine that the AI provider doubles its prices, restricts access to its application programming interface you built part of your business around, changes its terms of service or collapses. What do you do? A large corporation may absorb the disruption. A small business may not.

The small business may still have its customers and premises, but an increasingly important part of its productive capacity could suddenly be outside its control. The business owner has just realised that AI access is not the same thing as security.

AI is becoming part of how organisations operate, make decision and generate revenue, and businesses are increasingly being encouraged to integrate AI into important decisions and everyday operations

We spend a lot of time asking which country has the most advanced model, who is winning the AI race and whether China or the US will dominate the next technological era. Very important notes. But for the person whose small business depends on AI, equally important is can they trust this technology enough to build their businesses around it?

And, who is AI built for? Because once AI becomes part of someone's livelihood, that question becomes economic and deeply human.

This is why President Xi Jinping's remarks at the opening ceremony of the 2026 World AI Conference and High-Level Meeting on Global AI Governance in Shanghai on July 17 are worth examining.

President Xi said that countries should take a people-centred approach to AI, develop it ‘for the positive and for good,’ ensure that it contributes to shared prosperity and common security, and work together towards a just and equitable system of global AI governance.

If people are truly at the heart of what we do, their interests must not be disregarded once the product reaches the market. A people-centred approach to AI means that dignity, safety, trust and autonomy must be considered from the moment an AI system is designed, throughout its deployment and updates, and until it is eventually withdrawn.

Today's AI ecosystem spans closed, open-source and open-weight models. Closed models such as those developed by Anthropic and Open-AI are accessed through the company's platform, leaving pricing, updates and access under the provider's control.

Open-source models publish both the model and the code used to build it, allowing anyone to inspect and modify them, while open-weight models release the trained model, allow users to download and customise model weights within their own infrastructure.

For a small business, this difference is critical. Different AI ecosystems distribute power differently. Closed-model centralise control with the companies that own the models and the infrastructure behind them. Open-weight ecosystems create more opportunities for organisations to build on existing models without depending entirely on one provider.

That dependence is no longer hypothetical. In June 2026, the US government ordered Anthropic to suspend access to its newest frontier AI models for all foreign nationals, citing national security concerns.

Because Anthropic could not practically distinguish eligible and ineligible users across its global services, it temporarily disabled the models much more broadly while complying with the directive. Access was later restored after additional safeguards were introduced.

For businesses outside the United States, this showed that AI models could change overnight because of a geopolitical decision made in Washington rather than a technical failure or a commercial choice. Thus beyond innovation and competition, which AI ecosystem leaves people and businesses with greater control, resilience and trust when access itself can become a geopolitical decision?

The United States also illustrates how AI markets can become less people-centred when ownership, profits and decision making power around a technology, becomes highly concentrated by a small group even though millions of people use and contribute to it. A pattern political economists describe as elite capture.

The numbers speak for themselves. In 2026, California attracted $366 billion in venture capital, with closed AI model developers OpenAI and Anthropic alone accounting for roughly $217 billion of that investment through mega funding backed by Microsoft, Amazon, Nvidia and leading venture capital firms.

At the same time, Microsoft, Amazon, Google and Meta are projected to spend more than $700 billion on AI infrastructure in a single year, giving a handful of companies major control over the computing power needed to build frontier AI models.

The public participates in AI's growth primarily as users and data contributors, while the greatest financial returns go to shareholders, founders and infrastructure providers.

This divide is also reflected in national AI strategies. Washington's AI Action Plan prioritises maintaining US technological leadership through frontier AI models, advanced semiconductor controls and commercial AI platforms.

On the other hand, China through its Global AI Governance Initiative, places greater emphasis on open cooperation, technology sharing, open-weight model development and expanding AI participation among developing countries.

That is precisely where a people-centred AI approach differs. A people-centred AI ecosystem should be measured by whether citizens share in the value AI creates, shared prosperity, and not just from the millions of people using it.

The concentration of ownership helps explain why some corporations are increasingly investing in sovereign AI models. Just recently, on August 24, 2026, Thomson Reuters launched their own legal AI model Thomson-1.

The company built the model, using its own legal datasets facilitated by using Alibaba's open-weight Qwen model. Owning part of the AI stack would reduce long-term costs while giving it greater control over privacy, performance and specialised legal knowledge.

With such developments, there is hope for the Global South.

At the Shanghai conference, President Xi announced that China would provide developing countries with 5,000 AI training and seminar opportunities over five years, establish AI application cooperation centres with organisations including the African Union and BRICS, and enable 30 countries to use the AI-powered MAZU meteorological warning system

The African Union's Continental AI Strategy calls for an Africa-owned, people-centred, development-focused and inclusive AI ecosystem.
In Ghana, a new generation of AI founders are building healthcare, language and agricultural tools around AI. In Kenya, entrepreneur Kate Kallot's company Amini is building AI-powered environmental data infrastructure using African datasets instead of relying on imported climate models.

Kenya’s National AI Strategy 2025–2030 focuses on AI digital infrastructure, data and AI governance, and research, innovation and commercialisation. It also recognises data sovereignty, cybersecurity and ethical oversight as important to national interests.

The more Africa develops its own datasets, models, infrastructure and applications, the more capable it becomes in deciding which problems AI should solve and on whose terms. Still, Africa’s biggest constrain is electricity supply.

The International Energy Agency projects that global data-centre electricity demand could more than double by 2030 because of AI. Many African countries still struggle to provide reliable electricity for households and industries, making AI infrastructure a development challenge.

People-centred AI in Africa therefore begins with something as basic as reliable power, affordable connectivity and computing access. Without those foundations, AI risks widening the digital divide instead of closing it.

Still, the most important question is which model leaves ordinary users with greater capability, accessibility, resilience and trust.

Going back to the small business owner, the goal will be simply to keep the business alive, with a model that is dependable on Monday morning, secure enough to handle customer information, transparent enough to understand its limits and stable enough that tomorrow's software update does not destroy today's business model.

The writer is Project Coordinator, China Media Group Africa