The market behind the machine: Health AI, capital and the politics of dependence

“Technology never arrives on its own. It enters a system that already has its own interests, institutions and ideas about what matters.”

An artificial intelligence system can be seen as a way to improve efficiency in one country, a means of extending public services in another, a strategic capability in a third and a commercial opportunity somewhere else. The technology may be largely the same.

The circumstances around it are not. Who pays for healthcare, who regulates it, who owns the data, who provides the infrastructure and who ultimately decides what constitutes a good outcome can all change from one country to another.

Beyond Adoption: The State of AI in Healthcare Across the Global South

This matters because Health AI is gradually moving beyond individual applications. It is becoming part of how health systems
operate: helping interpret medical images, support clinical decisions, discover medicines, monitor populations and manage increasingly large amounts of information. Once that happens, the question of whether the technology is responsible cannot be separated entirely from the political and economic system in which it is being used.

There is a useful lesson here from Karl Polanyi, who wrote that the economy is embedded in social relations. Markets do not exist somewhere outside society; they operate through rules, institutions and choices made by people and governments. Healthcare perhaps makes this more obvious than any other sector.

Consider the role of private capital. Some of the world’s most important medical advances have depended upon companies willing to take risks, invest for years and develop products that governments alone might not have produced. The same is now true of Health AI. Investment is helping to develop diagnostic systems, drug discovery platforms and tools that could improve the delivery of care.

But investment naturally follows opportunity. A company has to ask whether a product can be developed, purchased and sustained. A public health system may ask a different question: whether the same product addresses an important need, including one for which there may be little commercial return. The history of medicines offers a useful reminder.

During the HIV/AIDS crisis, access to life-saving medicines became intertwined with questions of intellectual property and the ability of poorer countries to afford treatment. The debate eventually contributed to the 2001 Doha Declaration, which reaffirmed that intellectual property rules should be interpreted in a way that supports public health.

Health AI may produce a different version of the same question. If a model has been trained on vast amounts of health information, who owns the resulting capability? If a diagnostic tool is expensive to license, who gets to use it? And if an AI system becomes sufficiently useful that hospitals begin to depend upon it, does access remain simply a matter of purchasing power?

These are not arguments against markets. They are questions about what happens when something that begins as a commercial product gradually becomes part of essential infrastructure.

The picture looks different in a more state-led system. Governments that control large parts of healthcare can sometimes introduce technology at considerable scale. National procurement can bring together hospitals that would otherwise make separate decisions. Public investment can support infrastructure that may not have an immediate commercial return. A government can also coordinate regulation, procurement and implementation in a way that a fragmented market cannot easily replicate.

Who makes the decision

When the state is simultaneously the funder, regulator, purchaser and operator, implementation may become easier. Yet questions about consent, transparency and the ability of an individual to challenge a decision become questions of public governance as much as product design.

Amartya Sen’s idea of development as an expansion of people’s freedoms is useful here. Better healthcare is not only about delivering an outcome more efficiently. It is also about giving people greater ability to live the lives they value. An AI system that saves time but leaves a patient unable to understand or question an important decision would present a more complicated measure of progress than efficiency alone.

The Gulf offers yet another perspective. Countries such as the United Arab Emirates and Saudi Arabia have invested heavily in digital government, artificial intelligence and healthcare. Their highly coordinated systems can allow national priorities, infrastructure investment and technology programmes to move relatively quickly. That speed can be valuable. Building digital health infrastructure often requires decisions that cut across ministries, companies and public institutions, and a government with the ability to coordinate these actors has an obvious advantage.

But there is still a question worth asking: where does accountability sit when the same institution has considerable influence over investment, regulation and implementation? There is no universal answer. Different political systems simply distribute these responsibilities differently.

The more difficult question arises when technology crosses borders.

A health system in Africa or South Asia may use an AI model developed elsewhere, hosted on foreign cloud infrastructure and dependent upon computing hardware produced through international supply chains. The health data may come from local patients, while much of the underlying technology, capital and intellectual property sits outside the country.

The technology is accessible. But is the country able to shape it? This distinction between access and agency deserves greater attention.

The concern is not that technology companies are inherently problematic, nor that countries should attempt to develop every capability themselves. Tech investment and technological cooperation can bring enormous benefits. The concern is what happens when a country becomes dependent on capabilities over which it has little influence. There is a historical dimension to this.

Colonial economic relationships were shaped not only by the movement of goods, but also by who controlled knowledge, capital and production. Today’s digital economy is obviously not a continuation of colonial rule. Yet the underlying question – who creates value, who owns it and who has the power to make decisions about it – remains relevant.

For Health AI, this could become particularly important.

A country may provide the patients, the data and the clinical environment in which an AI system learns and proves its value, while the greatest economic value is captured elsewhere. It may adopt international standards without having had much influence over their development. It may gain access to advanced technology without developing the institutional capacity needed to evaluate or govern it. That is not necessarily exploitation. Sometimes it is simply how technological specialisation works.

A dependency worth understanding

This is also why responsible business cannot stop at regulatory compliance. A company entering a new healthcare market needs to understand more than whether its product meets a particular technical requirement. It needs to understand how healthcare is financed, how decisions are made, what infrastructure exists, how data is governed, what governments are trying to achieve and where responsibility ultimately sits.

Government affairs therefore also become more than a question of market access. It becomes part of understanding whether a technology can be introduced in a way that is useful, sustainable and appropriate to the country in which it operates.

The same applies to geopolitics. For years, discussions about strategic dependence in healthcare have focused largely on physical supply chains: medicines, medical devices, semiconductors and critical minerals. Those questions remain important. But Health AI adds another layer. Dependence can also sit in data, computing capacity, cloud infrastructure, models, technical expertise, capital and standards.

A country can have hospitals that it owns, doctors that it trains and patients that it serves, while still relying on foreign infrastructure for some of the intelligence increasingly embedded within its health system. This does not mean that every country needs technological self-sufficiency. That would neither be realistic nor necessarily desirable. It does mean that countries should have enough capability to understand what they are buying, assess whether it works in their circumstances, and retain meaningful influence over technologies that become important to their public services.

A technology that fits the institutional reality of a market is more likely to be adopted, trusted and sustained. For the Industry and companies, the implication is equally important. Responsible Health AI cannot mean simply asking whether a model works. It should also mean asking where it is being introduced, who will depend upon it, who benefits from it and whether the surrounding health system has the capacity to use it well.

Markets have a role in creating innovation. States have a role in ensuring public value. International cooperation can widen access. Local institutions provide the context in which technology acquires meaning. None of these approaches has a monopoly on good answers.