Artificial intelligence has left the pilot stage and moved into clinics, command centers, and policymaking tables across the WHO European Region. The World Health Organization’s new regional report, Artificial intelligence is reshaping health systems: state of readiness across the WHO European Region, is the first systematic scan of how 50 countries are planning, governing, financing, and actually deploying AI for care and public health. It reads as both a status dashboard and a route map: a careful balance of promise and risk, granular percentages that reveal where momentum is real, and pointed gaps that explain why adoption still stalls at the bedside or the budget desk (World Health Organization [WHO], 2025).
The document opens by situating AI within the region’s Digital Health Action Plan 2023–2030, underscoring equity, human rights, and resilient infrastructure as guardrails for any AI push. That framing matters. During COVID-19, countries with robust data governance and interoperable systems integrated AI faster into diagnostics and forecasting; others struggled with fragmentation, workforce gaps, and weak validation pipelines. The report’s through-line is that the same structural strengths and weaknesses still shape today’s AI rollout, only now the decisions are more consequential as tools shift from pilots to routine operations (WHO, 2025). In the Foreword, Regional Director Hans Kluge calls attention to a central tension: rapid diffusion of AI capabilities versus lagging clarity on safety, accountability, and inclusion—an imbalance that risks eroding trust if not corrected (WHO, 2025).
Methodologically, the report draws on a WHO-administered survey (June 2024 to March 2025) sent to all 53 Member States; 50 responded, yielding a 94 percent response rate. The unit of analysis is the health system, and the lens is deliberately wide: national strategies, legal frameworks, data governance, uptake patterns, stakeholder engagement, workforce readiness, private-sector roles, and adoption barriers. The analysis is mostly descriptive but bolstered with case sketches (for example, Slovakia’s AI-assisted radiotherapy planning procurement) that translate policy premises into operational choices. Country profiles accompany the main volume to surface subregional variation and EU-specific dynamics (WHO, 2025).
The heart of the report is organized into six thematic pillars. Each pillar functions as an analytic vignette with concise key figures and actionable signals.
The navigators—strategy and oversight—ask a simple question: who sets the course, with what instruments, and how coherent is the approach across government? Only 8 percent of responding countries have issued a health-specific AI strategy; 14 percent are developing one. Meanwhile, two-thirds report a national cross-sector AI strategy where health is either named or implied, and another third fold AI into their digital health strategies without a stand-alone health AI document (WHO, 2025). The governance model mirrors that hybridity: in 46 percent of cases, implementation sits with a single existing government body; in another 46 percent, responsibilities are split among several. A smaller share has created a new agency or an independent entity. This is the core trade-off the report keeps returning to: cross-sector strategies buy consistency and shared standards but can blur health priorities; health-specific strategies sharpen clinical alignment but can fragment regulation unless tightly coordinated (WHO, 2025).
The change-makers (stakeholder engagement and workforce)highlight a practical adoption truth: AI succeeds when the people who must use it are involved early and trained continuously. Seventy-two percent of countries reported some form of stakeholder consultation, but the mix is skewed. Focus groups and workshops are common, and government actors, clinicians, and developers dominate the conversation; patient associations and the broader public are the least engaged cohorts. Only 28 percent made consultation insights public—a missed opportunity for legitimacy and learning. On training, the pipeline is thin: 20 percent offer preservice education on AI, 24 percent provide in-service training, and 42 percent have created new roles for data or AI expertise inside health services (WHO, 2025). The report is explicit about the risk profile here: low digital and AI literacy at the front line drives automation bias, overreliance on tools, and erosion of clinical judgment; conversely, embedding AI literacy across professions creates capacity for critical appraisal, safe escalation, and responsible use (WHO, 2025).
The guardrails (law, policy, and guidance) capture the regulatory transition underway. Nearly half of countries have assessed legal gaps, and just over half report at least one regulator responsible for assessing and approving AI systems. But health-specific AI laws are rare; only 8 percent have formal liability standards tailored to AI in health, and a mere 6 percent have legal requirements that specifically address generative AI in care settings (WHO, 2025). The report’s argument is pragmatic: without clear liability allocation among developers, deployers, clinicians, and institutions, adoption will either stall from fear of risk or accelerate unsafely under ambiguous accountability. It also calls out the blurred border between regulated medical uses and consumer-grade wellness apps—an area where cross-border care and data usage complicate oversight and expose patients to variable safeguards (WHO, 2025).
The backbone (health data governance) tracks the infrastructure and rules that make AI safe and useful. Sixty-six percent of countries report a national health data strategy; 76 percent have or are developing a health data governance framework; and 66 percent have established a regional or national health data hub. These are strong macro signals, but the micro-rules that unlock value remain patchy: only 30 percent offer guidance for secondary use of health data for public-interest research and 30 percent have rules that enable cross-border data sharing for research. Without these missing pieces—consent models, transparency requirements, strong de-identification standards, and equity-minded access protocols—AI will either remain narrow and siloed or advance in ways that are opaque to the public (WHO, 2025). The report steers countries toward alignment with international norms, enhanced protections for vulnerable groups, and “good practice” networks for equitable data-hub design and rollout (WHO, 2025).
The catalysts translate aspirations into operations. Fifty-two percent of countries have named priority AI areas, but only a little more than half of those have earmarked funding. The top adoption drivers are tightly aligned with health-system stress points: improving patient care and outcomes (98 percent), reducing workforce pressure (92 percent), and increasing efficiency (90 percent). On the ground, AI-assisted diagnostics is the most common application (reported by 64 percent), followed by patient-facing chatbots (50 percent). The report threads a careful line: these tools can alleviate clinician workload and empower patients, but they also carry risks—biased outputs, performance drift, reduced clinician–patient interaction, and inequities for marginalized populations—if accreditation, pre- and post-deployment evaluation, and continuous monitoring are weak (WHO, 2025).
The gatekeepers (adoption barriers and enablers) name what stalls scale. The number-one barrier is legal uncertainty, reported by 86 percent of countries; the second is financial affordability at 78 percent. Asked which policy levers would unlock adoption, 92 percent pointed to clear liability rules and 90 percent to guidance on transparency, verifiability, and explainability (WHO, 2025). The report recommends regulatory sandboxes so regulators, developers, and providers can co-test systems under supervision; rigorous comparisons to non-AI alternatives; explicit human-rights checks; clarity on which responsibilities remain public; and transparent public–private partnership terms to protect community interests and ensure access to technology beyond vendor lock-in (WHO, 2025).
Beyond the numbers, the report functions as a synthesis of WHO’s normative guidance since 2021 on ethics, governance, and evidence standards for AI, including risk-based life-cycle regulation, clear intended-use documentation, external validation with independent datasets, prospective clinical validation for higher-risk tools, and rigorous post-market surveillance. It extends that guidance to large language models, flagging their communication and summarization potential alongside well-known pitfalls like hallucination and hidden bias—pitfalls that can be mitigated but not wished away (WHO, 2025).
The country examples interspersed through the findings translate governance choices into procurement criteria, workforce design, and change-management strategies. The Slovak radiotherapy case, for instance, combines qualitative vendor evaluation with measured efficiency targets (e.g., halving contouring time) and national scale considerations. That blend (fit-for-purpose metrics, equity in access, and public reporting) shows what “responsible acceleration” can look like when ministries, clinicians, and technical experts co-design from the outset (WHO, 2025).
The conclusion is neither techno-optimist nor defeatist. It argues for shared learning, regulatory alignment, and sustained, targeted investment as the only route to equitable, safe, people-centered AI—precisely because the hardest problems are not model architectures but governance choices. For ministries, the immediate to-do list is clear: articulate or update national strategies with time-bound objectives; close liability and transparency gaps; strengthen data-hub rules for secondary use and cross-border collaboration; mainstream AI literacy across preservice and in-service training; and attach real budgets to declared priorities. For providers, the work is to embed evaluation, bias monitoring, and escalation pathways into routine workflows so clinicians can contest or override machine output without friction. For the public, the pay-off must be visible in access, experience, and outcomes; that requires transparency by default, participatory rule-making, and oversight bodies with teeth (WHO, 2025).
As an integrated whole, the report’s argument is simple: AI can amplify health systems’ best instincts (equity, safety, efficiency, and solidarity) but only if countries do the slow institutional work that makes fast technology safe to trust. The data, examples, and specific percentages give policymakers and hospital leaders a practical baseline. What they do next will determine whether AI lightens the load for clinicians and patients or adds yet another layer of opaque complexity to systems already under strain (WHO, 2025).
References: World Health Organization. (2025). Artificial intelligence is reshaping health systems: State of readiness across the WHO European Region. Copenhagen: WHO Regional Office for Europe