Artificial intelligence is already being used in European hospitals faster than most countries are setting rules to govern it, said Hans Henri P. Kluge, the WHO Europe chief. Two-thirds of the 53 countries in the WHO Europe region were deploying AI diagnostics, and half had AI-powered patient chatbots. Only one in 12 had strategies for governing AI.
The figures focus attention on how health systems oversee technology introduced into clinical practice. Kluge called the situation “concerning” and warned that deploying AI without sufficient governance could put patients at risk and take a toll on human health. “All of this erodes public trust in health systems more broadly,” he said.
A governance gap across Europe
The figures show a gap between adoption and oversight across the WHO Europe countries. Two-thirds of the 53 countries used AI diagnostics, while only one in 12 had strategies for governing AI. Half had AI-powered patient chatbots. Kluge said European hospitals were deploying AI solutions, but only a few countries governed them.
The article doesn’t detail the rules used in individual countries or identify which countries have strategies. It says oversight needs to continue after an AI system has approval and enters clinical practice. A system’s use in a hospital, then, doesn’t mark the end of the oversight process described by the people interviewed.
HealthAI, the global agency for responsible AI, is working with regulators in 10 countries on a global early warning system. The system aims to identify adverse events and share information about emerging risks. Ricardo Baptista Leite, HealthAI’s CEO, said these mechanisms could help identify a problem in time and prevent harm at scale. “These kinds of mechanisms will allow us to avoid harm at scale, if something does go wrong, so that we can detect it in time and keep citizens safe,” he said.
Prevention, not just treatment
Baptista Leite said AI could help healthcare providers identify people at risk of certain conditions before they become ill, allowing earlier intervention. The approach would analyse health data alongside social factors that influence health, aiming to identify people who need support before their conditions deteriorate.
Europe’s ageing population is putting increasing pressure on healthcare systems that are already stretched. Baptista Leite said hospitals often receive payment to treat patients rather than keep them healthy. Moving toward prevention, he argued, would require funding models that reward early intervention and proactive monitoring.
He described the goal as preventing a heart attack rather than simply saving someone after one has happened. But AI alone wouldn’t shift healthcare from treatment to prevention. “If you retrofit AI into an inefficient system, it just makes you very efficient at being inefficient,” he told Euronews. He added: “Technology is not going to save us alone.”
Trust after approval
Scaling AI in healthcare depends on patients and healthcare professionals trusting that systems are safe and their benefits outweigh their risks. The article identifies continued oversight after approval as part of that requirement, alongside the early warning system HealthAI is developing with regulators in 10 countries.
Kluge’s warning connects the speed of deployment with the need for governance: AI is already being used for diagnostics and patient chatbots across the region, while formal strategies remain uncommon. The figures don’t establish that every system has caused harm. They do show the scale of adoption and the comparatively limited number of countries with AI-governance strategies.