A new line for clinical AI in Europe

A new line has been drawn for clinical AI in Europe

Across Europe, a new and more consequential conversation is taking place around the role of artificial intelligence in healthcare. In many settings, it is already part of the day-to-day clinical workflow — from drafting patient notes to supporting decisions where time is limited. Yet much of the early wave of AI slipped into practice without the expectations placed on new treatments, medical devices or clinical guidelines. That period is now giving way to a more structured regulatory environment.

This shift is one that Elsevier’s parent company, RELX, is engaged in directly. RELX has been appointed to the EU AI Act’s Advisory Forum — the official expert body advising the European Commission on implementation of the AI Act. Its appointment, from more than 700 applications and as one of only 32 companies, reflects the importance of ensuring that science, research, healthcare and trusted-information perspectives are represented as Europe defines responsible AI in practice.

Healthcare is not a neutral testing ground for innovation. Clinical AI operates in high-risk environments where patient safety, professional accountability and evidence integrity are non-negotiable. When AI tools influence clinical decisions, the margin for error narrows and the consequences of poor governance become far more serious.

Why care depends on a higher standard

Why patient care depends on holding healthcare AI to a higher standard

AI adoption in healthcare is accelerating faster than anyone predicted, but governance has lagged behind use. This imbalance is not sustainable, and it places an unreasonable burden on clinicians to compensate for gaps they did not create. Patient care is becoming more complex, with aging populations and increasing multimorbidity driving rising workloads — and when tools promise to save time, interest in them is only natural.

But the gap between what is governed and what is used in practice has a far greater impact in healthcare than in other industries. Clinicians train for years to hone judgment, assess evidence and understand where uncertainty sits, underpinned by a professional obligation to do no harm and act in the patient’s best interest.

If that gap is left unaddressed, reliance on tools that cannot be explained or examined will steadily erode that trust.

Patient safety cannot be an afterthought or a reaction to widespread implementation. Clinicians should not be expected to act as the final safeguard for tools that lack clear governance, validation or accountability. Responsible AI adoption cannot be achieved through individual vigilance alone — it must be embedded at the level of design, deployment and organizational oversight. When clinicians are already reaching for whatever is available, the task is to put credible, fit-for-purpose AI in their hands and make the safe, governed choice the easy one.

Ready for regulation — or not

Some clinical AIs are ready for regulation. Many are not.

Not all AI used in healthcare is created equal, yet clinical and generalist tools are too often discussed as if they carry the same risk and readiness. Many widely used, free generalist tools were not developed for clinical use: they are typically trained on broad internet content rather than curated medical evidence, and often lack the clinical oversight, governance and accountability expected in healthcare. As a result they can generate inaccurate, incomplete or outdated information and provide outputs with limited transparency around how conclusions were reached.

What should separate clinical AI from general-purpose tools is trust. The first question at the point of care is simple: where is the answer coming from, and can I trust it? Every answer should be explainable and verifiable, grounded in credible content a clinician can trace, check and challenge before acting on it. Systems that generate opaque responses from mixed or unverified sources are much harder to validate.

In healthcare, “free” or open access is not a governance model.

The next phase of clinical AI

Trust and leadership will define the next phase of clinical AI

AI will remain central to the future of healthcare innovation, but Europe is sending a clear message that scale alone is no longer enough. As regulatory expectations evolve, the EU AI Act reinforces principles healthcare has always depended on: patient safety, clinical accountability and evidence integrity.

Leadership in healthcare will not be defined simply by who adopts AI fastest, but by who adopts it responsibly and ethically.

Increasingly, that means giving clinicians a credible, evidence-led alternative and making it the easier choice, rather than leaving them to default to whatever is free and open. At the same time, patients now arrive with AI-generated information in hand, which makes trust even more important. For healthcare organizations, the challenge is selecting technologies that support clinical decision-making without undermining accountability, transparency or patient trust. The leaders who build on a credible, evidence-led foundation now will help define what trusted, responsible clinical AI looks like — anything less does not belong at the point of care.

Click to download the whitepaper and read the full contents.