WHO: European Hospitals Are Adopting AI Faster Than They Can Oversee It

WHO Regional Director for Europe Hans Kluge presented in Lisbon the results of the first review of this scale into the use of artificial intelligence across European health systems. The conclusion is simple: the technology is reaching hospitals far faster than countries are managing to build rules around it.

The data comes from a WHO/Europe survey covering all 53 countries in the region, from Portugal to Central Asia. The report shows AI entering hospitals mainly through imaging diagnostics, X-ray and CT scan analysis, clinical decision-support systems, and automated patient communication.

Gap Between Deployment and Oversight

Kluge called this gap the central problem of the current phase of health digitization. Nearly every country cites improved patient care as the main reason for adopting AI, but fewer than half have even checked whether their existing legal frameworks are fit for regulating the technology.

This gap between deployment and governance is now the biggest challenge for AI in health - Hans Kluge, WHO Regional Director for Europe

In practice, this means diagnostic tools built on machine learning models are reaching doctors' offices without unified rules on liability for errors, without requirements for testing across diverse patient populations, and without a common standard for algorithmic transparency.

Risk to Patients

Kluge warned directly about the clinical consequences of malfunctioning systems. The concern isn't just technical mistakes, but systemic bias in algorithms that, trained on incomplete or unrepresentative data, may be worse at detecting disease in certain patient groups.

A faulty algorithm can deliver the wrong diagnosis, to a real patient, with real consequences - Hans Kluge, WHO Regional Director for Europe

As an example of a well-implemented tool, WHO pointed to a hospital in Coimbra, Portugal, where an AI system supports image analysis for detecting chest conditions and bone fractures, shortening queues in emergency and primary care. The report notes, however, that such deployments remain isolated projects in most countries rather than part of a coherent national policy.

Staffing and Training Gaps

The biggest weakness turns out to be workforce preparation. Only one in five countries teaches future doctors and nurses to use AI tools during their studies, and just one in four offers training to staff already in the profession. Without this, medical personnel often end up using AI systems through trial and error, without understanding their limitations.

WHO announced it will launch a regional roadmap on AI and health by 2028, aimed at standardizing safety assessment criteria, requirements for training data, and rules on legal liability for algorithm-supported decisions. At the Lisbon meeting, representatives from 37 countries across six WHO regions also proposed a separate Lisbon initiative for cooperation among Portuguese-speaking countries in this field, with a planned launch at a health summit in Brazil in 2028.

Implications for Poland and the EU

The report lands as the EU's AI Act begins to cover high-risk systems, a category that includes most AI-based medical tools. Polish hospitals, including a facility in Wadowice that is deploying AI to read X-ray and CT scans, are testing solutions similar to those described in the WHO report, but a national strategy for overseeing this technology remains in its infancy.

For Polish hospital directors and officials responsible for implementing the AI Act, the WHO data serves as a warning: buying and switching on a diagnostic tool is only the beginning. Without quality-assessment procedures, staff training, and clear rules on liability for a system's faulty recommendation, the benefits of AI risk being overshadowed by lost patient trust and legal risk for institutions.