Artificial intelligence is rapidly becoming part of medical research, but the growing use of AI in health studies is creating new questions about privacy, transparency, accountability and patient protection.

The World Health Organization is highlighting the need for stronger ethical oversight of research involving artificial intelligence, with particular attention to how health-related AI studies are reviewed before and during their implementation.

The issue has become increasingly important as researchers use AI to analyse medical records, identify disease patterns, assist diagnosis, develop medicines and process enormous amounts of biomedical data.

While these technologies can accelerate scientific work, their use also introduces risks that traditional research oversight systems may not always be designed to address.

AI Is Changing Medical Research

Artificial intelligence can process huge quantities of information much faster than conventional analytical methods.

Researchers can use machine-learning systems to identify patterns in medical images, examine genetic information, analyse clinical records and explore potential relationships between diseases and treatments.

AI can also help researchers identify promising drug candidates and organize information from large scientific databases.

These capabilities could potentially shorten parts of the research process.

However, speed and computational power do not automatically guarantee accurate or ethical results.

An AI system can produce misleading conclusions if the data used to train or evaluate it are incomplete, biased or poorly representative.

Health Data Requires Special Protection

Medical research frequently involves highly sensitive information.

Patient records can contain details about illnesses, medications, genetic characteristics, treatments and other personal information.

When such data are processed by AI systems, researchers must consider how information is collected, stored, transferred and protected.

A data breach involving medical information can have serious consequences for individuals.

There is also a risk that information believed to be anonymized could potentially be combined with other datasets to identify individuals.

For this reason, AI-related health research requires careful attention to data governance and privacy protection.

Traditional Ethics Review May Not Cover Every AI Risk

Research ethics committees already evaluate studies involving human participants.

They commonly examine issues such as informed consent, potential harms, benefits, privacy and participant safety.

AI can introduce additional questions.

For example, an algorithm may change as it receives new data or may behave differently when deployed in a population that was poorly represented in its original dataset.

Researchers may also find it difficult to explain precisely why a complex machine-learning model generated a particular result.

These characteristics can create challenges for conventional oversight frameworks.

Transparency Is Becoming More Important

A major issue surrounding AI research is transparency.

Researchers and participants may need to understand what type of AI system is being used, what information it processes and how its outputs influence research decisions.

Complete technical explanations may not always be possible for highly complex models, but this does not eliminate the need for meaningful accountability.

Research institutions can establish documentation requirements that explain the purpose of an AI system, its intended use, limitations and potential risks.

Clear records can also make it easier to investigate problems if an AI-supported study produces unexpected results.

Bias Can Enter Through Training Data

AI systems learn patterns from data.

If the data contain historical inequalities or underrepresent certain populations, an algorithm can reproduce or amplify those patterns.

For example, a medical AI system trained mainly on data from one population may perform differently when applied to people with different genetic, demographic or socioeconomic characteristics.

This can become an important ethical issue if an AI system influences clinical research or future medical decisions.

Researchers therefore need to examine whether study data adequately represent the populations for whom the results may eventually be used.

Consent Becomes More Complicated

In conventional research, participants may be told how their information will be used.

AI can complicate this process because data may subsequently be used for purposes that were not fully anticipated when the information was originally collected.

Researchers may combine datasets or develop new analytical methods that were not available when participants initially provided their information.

This raises questions about whether existing consent arrangements remain appropriate as technology evolves.

Ethics oversight can help determine when additional safeguards or updated consent processes may be necessary.

AI Can Support Drug Discovery

Despite these concerns, AI also offers potentially valuable opportunities for medicine.

Researchers are using AI to examine molecular structures, predict interactions between compounds and biological targets, identify possible drug candidates and prioritize laboratory experiments.

These systems can help scientists evaluate large numbers of possibilities that would be difficult to investigate manually.

But AI-generated predictions still require experimental validation.

A computer model can identify a promising candidate, but laboratory and clinical research are needed to establish whether a treatment is actually safe and effective.

Medical Imaging Is Another Major Application

AI is increasingly being studied for applications involving medical images.

Algorithms can be trained to detect patterns in X-rays, CT scans, MRI images and other forms of diagnostic data.

In research settings, such systems can help scientists identify subtle patterns that may warrant further investigation.

However, performance can vary depending on the quality of images and the population represented in the training data.

A model that performs well in one hospital may not necessarily produce identical results elsewhere.

This makes independent validation an important part of responsible AI research.

Human Oversight Remains Essential

AI should not automatically replace human judgment in medical research.

Researchers and healthcare professionals remain responsible for evaluating whether an AI-generated result is scientifically credible and clinically meaningful.

Human oversight is especially important when an AI system produces an unexpected result or operates outside the conditions for which it was validated.

The role of human experts may also change as AI capabilities improve, shifting from performing every analytical task manually toward supervising, validating and interpreting AI-supported processes.

Cybersecurity Is Part of AI Ethics

Ethical oversight also needs to consider cybersecurity.

Medical AI systems can become targets for cyberattacks because they may contain valuable health information or influence research processes.

Attackers could potentially attempt to steal sensitive data, manipulate datasets or interfere with AI systems.

Protecting AI-supported research therefore requires technical security measures alongside ethical review.

Data encryption, access controls, monitoring and secure infrastructure can all contribute to reducing these risks.

Developing Countries Need a Voice

The development of global AI health standards also raises questions about representation.

Much AI research and development is concentrated in technologically advanced economies.

If global standards are created without sufficient input from lower- and middle-income countries, they may fail to address local healthcare conditions and resource limitations.

Different countries may have different legal systems, health infrastructures and cultural expectations surrounding medical data.

International cooperation therefore needs to include a broad range of stakeholders.

Regulation Must Keep Pace With Technology

AI systems are developing rapidly.

Regulatory and ethics frameworks can take considerably longer to create and implement.

This creates a risk that researchers and institutions may encounter technologies for which existing guidelines provide limited direction.

Flexible frameworks may therefore be needed.

Such frameworks should be capable of adapting as AI systems become more sophisticated while maintaining core principles such as safety, privacy, fairness, transparency and accountability.

Researchers Need Clear Guidance

Researchers can face uncertainty when deciding how to use AI responsibly.

Clear institutional policies can help define which AI applications require additional review, how datasets should be documented, what information participants should receive and how AI-generated results should be validated.

Training is also important.

Scientists, ethics committee members and health professionals may need to understand both the capabilities and limitations of modern AI systems.

Technical expertise and ethical expertise need to work together rather than operating separately.

The Balance Between Innovation and Protection

The central challenge is finding a balance between encouraging useful innovation and protecting people from avoidable risks.

Overly restrictive rules could slow potentially valuable research.

Insufficient oversight could allow privacy violations, biased results or unsafe practices to enter medical research.

A proportionate approach can allow researchers to benefit from AI while ensuring that higher-risk applications receive stronger scrutiny.

The level of oversight may therefore need to reflect factors such as the sensitivity of the data, involvement of human participants and potential consequences of an incorrect AI output.

A New Era for Health Research

Artificial intelligence has the potential to transform how medical research is conducted.

It can process information at unprecedented scale, assist scientific discovery and help researchers investigate questions that were previously difficult to address.

But the technology also changes the nature of research risks.

Privacy, bias, transparency, accountability and cybersecurity must therefore become central parts of AI-related health research rather than afterthoughts.

The WHO’s focus on ethical oversight reflects the growing recognition that technological progress and responsible governance need to develop together.

As AI becomes increasingly integrated into medicine, the quality of the technology will not be the only measure of success. The way it is developed, tested, governed and used will also determine whether its benefits can be realized without compromising public trust and the rights of people whose health information makes that research possible.