Artificial intelligence is reshaping health systems: state of readiness across the WHO European Region

Artificial intelligence is reshaping health systems: state of readiness across the WHO European Region — WHO/Europe, 2025

This report presents the first assessment of AI integration into health systems across the whole of the WHO European Region, based on findings from the 2024–2025 survey on AI for health care.

— the report’s abstract

TL;DR

WHO/Europe’s first region-wide survey of national AI-for-health policy. The unit is the country, not the hospital: 50 of the 53 Member States answered, through government-nominated coordinators. It measures readiness (strategies, law, data governance, workforce), not how much AI hospitals use.

1. Few health-specific AI strategies. 4 of 50 (8%) have published a national health-specific AI strategy; 7 (14%) are developing one. 33 (66%) have a cross-sector national AI strategy (pp. x, 9).
2. Liability is unsettled. 4 of 50 (8%) have developed liability standards for AI in health; 3 (6%) have legal requirements for generative AI in health care (p. xii).
3. Legal uncertainty is the top barrier (43 of 50, 86%), then financial affordability (39 of 50, 78%). Clear liability rules are the most-cited enabler (46 of 50, 92%) (p. xv).
4. Workforce training is thin. 12 of 50 (24%) offer in-service AI training and 10 (20%) pre-service training; 21 (42%) have created new AI or data-science roles (p. xi).
5. Use is reported at country level. 32 of 50 (64%) report AI-assisted diagnostics and 25 (50%) chatbots for patient support. 26 (52%) have named priority areas for AI, and “only just over half of these had allocated funding” (p. xiv).

Key data

Measure	Member States	Share	Page
Responded (of 53)	50	94% response rate	ix, 6
Published national health-specific AI strategy	4	8%	x, 9
… developing one	7	14%	x, 9
Health covered in a cross-sector AI strategy / AI in the digital health strategy / none	16 / 17 / 6	32% / 34% / 12%	9
National cross-sector AI strategy	33	66%	x, 10
Engaged stakeholders on AI in health	36	72%	xi
In-service AI training / pre-service AI training	12 / 10	24% / 20%	xi
New professional roles for AI and data science	21	42%	xi
Assessed legal gaps for AI in health	23	46%	xii
Liability standards for AI in health	4	8%	xii
Legal requirements for generative AI in health care	3	6%	xii
Regulatory agencies to assess and approve AI systems	27	54%	xii
National health data strategy / health data hub	33 / 33	66% / 66%	xiii
Guidance on secondary use of health data	15	30%	xiii
Identified priority areas for AI in health	26	52%	xiv
AI-assisted diagnostics in use / chatbots for patient support	32 / 25	64% / 50%	xiv
Top barrier: legal uncertainty / second: financial affordability	43 / 39	86% / 78%	xv
Enabler: liability rules / transparency and explainability guidance	46 / 45	92% / 90%	xv

Method and definitions

• Fieldwork. June 2024 to March 2025, online or on paper, in English and Russian (p. 6).
• Who answered. Each Member State was asked to nominate a national survey coordinator, who gathered the views of national digital-health and AI experts. One response per state. The UK response covers England only. Bosnia and Herzegovina, Monaco and Turkmenistan did not take part (p. 6).
• Definitions. A national AI strategy in health is “a high-level, government-endorsed document or framework” for AI in the health sector (p. 9). Some questions were asked only of states that answered yes to a prior question, so some shares have smaller bases (p. 6).
• Funding. Co-funded by the European Union. External peer reviewers include three European Commission staff.

Scope and reliability

• Country-level self-report by government respondents. The authors name the limits: broad terminology “allowing for country-specific interpretations”, dependence on the coordinators’ expertise, and findings that “may outdate quickly” (p. 7). They claim “documentary corroboration of submissions” as a strength.
• Not a measure of hospital adoption. “64% report AI-assisted diagnostics” means a state reports such use somewhere in its system, not that 64% of hospitals use it.
• No per-country figures in the main report. Country profiles are published separately and were not fetched. The four states with a health-specific AI strategy are not named in the sections read.

Dynamic-capabilities reading

• digital-sensing/digital-scouting: a scan of the policy environment health organisations in 50 European countries work in, covering strategy, law, liability, data governance and workforce training.

How it connects

• responsible-ai — liability standards, transparency guidance and post-market monitoring at national level; most states have not set them.
• enterprise-ai-adoption — context for the hospital-level figures: the rules hospitals deploy under are mostly unsettled.
• ai-maturity-measurement-comparison — a readiness measure at country level, a different unit from the organisational and worker measures in the thread.
• In this batch: European Commission deployment study surveys hospitals and clinicians in the same region and also finds unclear liability among the leading barriers.
• ai-in-health-care: the concept page that sets the health-care sources in this wiki side by side.

What was actually ingested

The full 76-page PDF from WHO IRIS. Read closely: foreword, executive summary, methods and limitations, and section 3.1 on national strategies. Sections 3.2–3.6 were searched for the figures above but not read in full. Bar charts are flattened in the text conversion; the infographic percentages and counts survive.

Linked entities and concepts

• responsible-ai, enterprise-ai-adoption — concepts. ai-maturity-measurement-comparison — thread.

Dangling (single-source mention, deferred): WHO Regional Office for Europe (publisher; no entity page), Nicholas Letchford, Hanna Dunning, Nagui Salama (WHO consultants, main authors), Keyrellous Adib (WHO/Europe, main author), Natasha Azzopardi-Muscat and David Novillo Ortiz (technical conceptualization), Hans Henri P. Kluge (Regional Director, foreword).