Fifty-three countries in the WHO European Region have unanimously adopted the region’s first roadmap for artificial intelligence in health. Covering 2026–2033, it does not treat AI as a shortcut to better care. Instead, it frames adoption as a health-system task that must combine useful technology with governance, rights, reliable data and a prepared workforce.
What the roadmap is—and is not
The roadmap was approved at the 76th session of the WHO Regional Committee for Europe. WHO says AI may help identify disease earlier, alert teams when a patient is deteriorating, reduce administrative burden and improve access to basic information. Those are possible uses, not guaranteed outcomes. Whether a tool helps depends on its evidence, the setting, the quality of implementation and continued human accountability.
Test 1: Is there clear governance?
The first priority is governance and leadership. Health organizations need named responsibilities, decision rules and a way to stop or modify a system when performance or safety falls short. A tool should not enter routine care simply because it appears efficient in a demonstration. Leaders need to define the clinical problem, the intended users, the affected population and the evidence required before deployment.
Test 2: Are rights and safety protected?
The second priority covers ethical, legal and regulatory safeguards. Patients should know when AI materially contributes to their care, while clinicians and institutions remain accountable for decisions. Procurement and deployment should consider bias, accessibility, privacy, security and avenues for review. A high average performance number can conceal weaker results for a subgroup, so evaluation must look beyond a single headline metric.
Test 3: Can the data foundation be trusted?
WHO’s third priority is data and digital infrastructure. Useful models require appropriate, high-quality data, interoperable systems, secure access and cybersecurity. Health data reflect real-world gaps and unequal access to care; they are not automatically neutral. Teams therefore need to understand where data came from, what is missing and whether the population using a tool resembles the population in which it was evaluated.
Test 4: Can people use and evaluate it?
The fourth priority is capacity-building, implementation support and learning. Training must extend beyond operating an interface. Health workers need to understand a system’s intended use, limitations and escalation path. Organizations also need real-world monitoring: whether the tool improves a meaningful outcome, creates extra work, changes access or produces new safety signals after launch.
A regional framework, local evidence
WHO drew on a region-wide readiness assessment involving responses from 50 of the region’s 53 countries. The roadmap aligns with European and international instruments, including the EU AI Act, the European Health Data Space and medical-device rules. It is accompanied by an implementation guide intended for countries at different levels of digital maturity.
Alignment does not remove the need for local review. A national strategy cannot establish that every product is safe, effective or appropriate for a particular clinic. Each use still needs evidence, transparent procurement, clinical oversight and evaluation in the population and workflow where it will operate.
What patients and professionals can ask
Practical questions include: What decision does the system support? What happens when it is wrong? Who can challenge its output? Is performance monitored across patient groups? Does it reduce workload or merely move it? Is there a non-digital route for people who cannot use the tool? These questions turn responsible AI from a slogan into an operational standard.
Source: WHO Europe, 29 October 2026. This article is general information and does not provide medical advice.
