The pulse of progress: AI, real-world data, and the ethics of continuous monitoring in healthcare
Artificial intelligence is rapidly reshaping how people interact with healthcare information. The emergence of large language models (LLMs) capable of interpreting complex medical knowledge has accelerated a shift that many in digital health have anticipated for years: the gradual breakdown of barriers between patients and the data that defines their health.
For the first time, individuals can query sophisticated systems about symptoms, therapies, or disease risks and receive immediate, structured answers. When combined with the growing ecosystem of wearable devices and continuous monitoring technologies, these tools have the potential to usher in a new paradigm of patient empowerment. Health information is no longer confined to hospitals, laboratories, or physicians' offices. Instead, it is increasingly accessible at your fingertips.
Yet, as healthcare enters this new era of AI-enabled insight, it is becoming clear that the technological promise must be matched by thoughtful guardrails. Without them, the same tools designed to democratise health data could undermine patient trust or even create new risks for public health systems.
At the centre of this debate is the simple question – what kind of data should train and power the next generation of healthcare AI?
The foundation problem: Data quality
Artificial intelligence systems are only as reliable as the information used to train them. While consumer-facing chatbots may appear authoritative, their recommendations ultimately reflect patterns learned from underlying datasets.
In many cases today, those datasets include large volumes of publicly available internet information. In consumer applications, this may be acceptable. In medicine, it is not.
A clinical recommendation carries far more weight than a restaurant suggestion or travel itinerary. If a model gives equal consideration to rigorous peer-reviewed research and anecdotal commentary from an online forum, the result is not simply “noisy” data; it may be clinically dangerous.
Healthcare AI will therefore require a new standard of validated training data, built on structured clinical evidence and scientifically verified sources. That foundation must include not only traditional medical literature, but also dynamic data reflecting how the human body changes over time.
This is where advances in biosensing and continuous monitoring are beginning to play an important role. Technologies capable of measuring physiological signals or molecular markers continuously can provide a far more nuanced picture of human biology than periodic clinical snapshots. Instead of seeing health as a sequence of isolated events, blood tests, scans, and diagnoses, medicine can begin to observe the body as a constantly evolving system.
For AI developers, this kind of longitudinal data could represent a transformative shift. Continuous datasets allow algorithms to learn not only what a disease looks like at a specific moment, but how it develops, responds to therapy, or resolves over time.
In other words, the future of healthcare AI may depend less on the volume of data available and more on its biological fidelity.
Europe’s regulatory moment
The debate around health data is unfolding at a particularly significant moment in Europe. Policymakers across the European Union are working to build frameworks that balance innovation with patient rights, including the proposed European Health Data Space (EHDS).
The initiative aims to create interoperable infrastructure allowing health information to be shared securely across EU member states for both clinical care and research. If implemented effectively, it could provide researchers and developers with access to unprecedented datasets spanning diverse populations and healthcare systems.
At the same time, Europe remains one of the most stringent environments for data protection. Regulations such as the General Data Protection Regulation (GDPR) have established global benchmarks for consent, transparency, and patient control over personal information.
For digital health innovators, these developments present both opportunity and responsibility. The EHDS could dramatically accelerate biomedical research and AI development by enabling ethically governed access to large-scale health datasets. But it will also require new approaches to data stewardship, governance, and transparency.
The European conversation highlights an essential reality: health data is not just a technical resource; it is a social contract.
Ownership, consent, and trust
Healthcare systems rely fundamentally on trust. Patients share deeply personal biological information because they believe it will be used responsibly and in their best interest.
Artificial intelligence complicates this relationship. Models trained on large datasets may derive insights from thousands or millions of patient experiences. But many individuals may not fully understand how their data contributes to these systems.
This raises difficult questions. If an AI tool recommends a therapy based on patterns learned from past patient outcomes, were those patients aware their data might inform such systems? Did they provide explicit consent for that use? And do they retain the right to withdraw their data from training sets?
As data aggregation accelerates, maintaining transparent consent frameworks will be essential. Otherwise, the drive toward “big data” risks becoming an opaque process that extracts value from personal health information without adequate accountability.
Europe’s regulatory landscape may offer useful guidance here. By emphasising patient rights, informed consent, and cross-border governance, European policymakers are effectively testing how large-scale health data ecosystems can function without compromising individual autonomy.
AI advice and the problem of authority
Another challenge lies in the way AI presents medical information to users.
Large language models are designed to communicate with confidence and fluency. For many patients, this creates the impression that the system possesses clinical authority. Yet, these tools remain probabilistic models, rather than medical professionals.
This dynamic becomes particularly concerning when individuals use AI systems without professional supervision. Patients facing complex diagnoses or chronic conditions may rely heavily on automated advice without understanding its limitations.
To mitigate this risk, healthcare AI systems will likely need validation layers that cross-reference outputs against established clinical guidelines and regulatory frameworks before recommendations reach users.
In other words, AI should function less as an autonomous advisor and more as a decision-support system grounded in verified medical standards.
Privacy in the era of predictive health
Data privacy concerns also become more complex as health technologies grow more predictive.
Consider a scenario in which an individual uploads an image of a skin lesion to an AI tool or searches repeatedly for symptoms associated with melanoma. That information could reveal potential health risks long before a formal diagnosis exists.
If such behavioural data were accessed by third parties, whether insurers, employers, or data brokers, it could theoretically influence decisions about coverage, employment, or financial risk assessment.
The implications extend beyond traditional data breaches. The greater concern may lie in the secondary uses of health-related data trails that individuals leave across digital platforms.
As predictive medicine advances, policymakers will need to ensure that emerging technologies do not inadvertently create new forms of discrimination or surveillance tied to health status.
Building a responsible future
Despite these challenges, the potential benefits of AI-enabled healthcare remain enormous. Continuous monitoring technologies, advanced biosensors, and intelligent analytics may eventually allow clinicians and patients to detect disease earlier, personalise therapies more precisely, and monitor recovery in real time.
Achieving that future will depend not only on technological progress, but also on ethical infrastructure. Developers, clinicians, regulators, and patients must work together to define how health data is collected, shared, and interpreted.
Three priorities stand out.
First, healthcare AI must be built on high-quality, clinically validated datasets, rather than unfiltered internet information.
Second, data ecosystems must prioritise transparent consent and patient ownership to maintain trust in digital health innovation.
Third, AI systems should incorporate robust clinical validation layers, ensuring that automated insights align with established medical standards.
If these safeguards are implemented thoughtfully, artificial intelligence could become one of the most powerful tools ever developed for improving human health. But if they are ignored, the technology risks undermining the very trust upon which medicine depends.
The promise of AI in healthcare is not simply about faster answers or larger datasets. It is about achieving a clearer understanding of the body’s complex signals and using that understanding responsibly to make informed decisions.
As the next generation of health technologies emerges, the challenge will not be whether we can build these systems, but whether we can guide them with the caution, ethics, and scientific rigour that medicine demands.
About the author
Dr Ben Hwang has served as Profusa’s Chairman of the Board and CEO since January 2012. Prior to Profusa, Dr Hwang served in a variety of leadership roles at Life Technologies Corp (acquired by Thermo Fisher Scientific, Inc.), including president of the Asia Pacific region and head of the qPCR division. Prior to joining Life Technology, Dr Hwang was a consultant with McKinsey & Company. Dr Hwang received his MA in Biology and PhD in Biology from The Johns Hopkins University.
