When public health becomes predictive

For much of modern history, healthcare systems have functioned reactively. A patient falls ill, symptoms emerge, a diagnosis is made, and treatment follows. Public health, meanwhile, operated at population scale through vaccination campaigns, disease surveillance, and preventive programmes designed to reduce broad societal risk.

Artificial intelligence is beginning to alter that relationship. Increasingly, healthcare systems are using AI not only to treat illness, but to predict outbreaks, identify high-risk populations, anticipate healthcare utilisation, and support earlier interventions. The potential benefits are considerable, particularly in overstretched systems where predictive infrastructure may strengthen prevention, improve resource allocation, and enable more responsive care.

Yet these developments introduce a deeper governance question: what happens when healthcare systems evolve from treating illness towards continuously inferring behavioural and biological risk across entire populations?

The quiet expansion of institutional visibility

Modern healthcare systems are no longer built solely around hospitals and clinics. Increasingly, they operate through data infrastructures.

Health records, diagnostic systems, pharmacy databases, insurance platforms, wearable devices, mobility information, social determinants data, and behavioural interactions now collectively generate a continuous flow of information about how individuals live, move, consume, age, and interact with healthcare systems.

AI changes the significance of this data.

Traditional public-health systems relied heavily on aggregated patterns and retrospective analysis. AI systems increasingly allow institutions to move toward probabilistic inference at scale-identifying not simply what populations are experiencing now, but what individuals and groups are statistically likely to experience in the future.

Recent developments across ASEAN illustrate this transition. At the 48th ASEAN Summit, leaders reaffirmed commitments to strengthen disease surveillance, operationalise the ASEAN Centre for Public Health Emergencies and Emerging Diseases (ACPHEED), and expand the use of digital health systems and AI across healthcare delivery and public-health management. These initiatives are intended to improve preparedness and resilience. Yet they also reflect a broader shift within public health: from responding to crises after they emerge towards anticipating vulnerabilities before they fully materialise.

This distinction matters because prediction gradually changes the relationship between institutions and citizens. A healthcare system responding to disease operates differently from a healthcare system continuously modelling behavioural probability.

From treatment to anticipatory governance

This is where the ethical conversation becomes more complicated than privacy alone.

Much of today’s AI governance discourse focuses understandably on data protection and consent. These are important safeguards. But the larger structural transformation concerns how predictive systems reshape institutional power itself.

Risk prediction models may increasingly influence: insurance eligibility; prioritisation of healthcare access; preventive interventions; public-health targeting; allocation of social support; and behavioural recommendations delivered through digital platforms.

Often, these processes occur through algorithmic systems that remain only partially visible to the individuals being assessed.

Over time, healthcare may therefore begin shifting from episodic care toward a more continuous form of anticipatory governance-where institutions increasingly intervene not only based on illness, but based on inferred future probability.

This does not necessarily emerge through coercion. In many cases, it emerges through optimisation.

This evolution is increasingly visible within healthcare institutions themselves. In Western Australia, Royal Perth Hospital recently announced plans to trial AI systems designed to forecast patient flow, identify individuals who may require additional support services, and assist hospitals in planning earlier interventions around discharge and resource allocation.

Importantly, the initiative is framed not as a replacement for clinical judgement, but as a mechanism to improve efficiency within an overstretched health system. Nevertheless, it illustrates how predictive systems increasingly shape institutional responses before problems fully manifest.

A public-health authority seeking to reduce cardiovascular disease burden may identify individuals considered statistically vulnerable and increase behavioural nudges around diet, physical activity, or medication adherence. Insurers may incentivise “healthy” behavioural patterns through wearable-linked systems. Healthcare platforms may personalise recommendations based on behavioural prediction models.

Individually, many of these interventions may appear beneficial.

Collectively, however, they gradually expand the capacity of institutions to shape behaviour through continuous observation and probabilistic classification. The line between healthcare and behavioural governance becomes less distinct.

Healthcare systems were never entirely neutral

None of this means public health itself is inherently problematic.

Public-health systems have always involved some degree of societal coordination and behavioural influence. Vaccination campaigns, tobacco control measures, sanitation policy, and infectious disease containment all required balancing individual autonomy with collective welfare.

But AI introduces scale, granularity, and continuity at levels historically unavailable to institutions.

The difference is not merely that systems become more data-rich. It is that they become increasingly capable of producing highly individualised forms of inference across entire populations simultaneously.

Importantly, these systems do not operate independently of social and political context.

A predictive public-health system operating within a highly trusted democratic environment may function very differently from one deployed in settings characterised by weak accountability, fragmented governance, limited transparency, or significant institutional asymmetries.

This becomes particularly significant across parts of the Global South, where digital public infrastructure is often expanding more rapidly than the governance capacity surrounding it.

In many countries, healthcare digitisation is occurring alongside uneven regulatory systems; weak interoperability standards; limited public awareness around data rights; and increasing dependence on private technology ecosystems.

Under such conditions, predictive public-health infrastructure may gradually evolve faster than the democratic safeguards required to supervise it meaningfully.

AI governance must account for sustainability, not just safety

The risk is not alone, but institutional drift

The concern is therefore not that public-health AI systems are inherently harmful.

In many cases, they may substantially improve healthcare outcomes.

The deeper concern is that societies may gradually normalise forms of institutional visibility and behavioural inference whose long-term implications remain insufficiently debated publicly.

Because once prediction becomes embedded into healthcare systems, institutional incentives also begin changing.

Healthcare systems may increasingly prioritise measurable behavioural optimisation. Insurance ecosystems may rely more heavily on predictive classification. Governments may become more comfortable governing through data-mediated intervention. Citizens may increasingly experience healthcare systems not primarily through human relationships, but through algorithmic assessment and behavioural scoring infrastructures.

None of these shifts happen suddenly.

Indeed, many emerge through initiatives explicitly designed to strengthen public health, improve operational efficiency, or enhance preparedness. That is precisely why they warrant careful democratic scrutiny: the most consequential transformations in governance often occur incrementally through systems introduced with broadly beneficial intentions.

That is precisely why governance becomes so important.

Responsible public health requires democratic limits

The challenge ahead is therefore not whether societies should use AI within public health. They almost certainly will.

The real question is whether democratic societies can establish meaningful limits around inference, prediction, and institutional intervention before these systems become deeply normalised infrastructure.

This requires moving beyond narrow conversations around technical compliance alone.

Questions of public-health AI governance increasingly involve: transparency around predictive systems; meaningful human oversight; institutional accountability; limits on behavioural manipulation; safeguards against discriminatory classification; and clear mechanisms for contestability and redress.

Equally important, societies may need to preserve a distinction between supporting public health and governing populations primarily through anticipatory behavioural logic.

Beyond data collection

As AI becomes more deeply embedded into healthcare governance, the defining question ahead may no longer be how much health data societies can collect. It may be how societies define the legitimate boundaries of inference itself.

Because eventually, the central challenge may not concern whether institutions possess the technical capability to predict behaviour and health risk at scale, but whether democratic societies retain the political and ethical capacity to decide when prediction should stop.