Scaling artificial intelligence in health
A · Strong evidence An OECD Health Division report backed by a member-country survey and an expert group, solid institutional evidence but a sector policy checklist rather than a field-shaping flagship.
Summary based on the full text·Original: about 100 minutes to read
In short
An OECD report (April 2026, revised edition) by Eric Sutherland, Rachel Fellner and Yunona L'Heureux of the OECD Health Division, funded by the Health Foundation. It finds AI is used in health administration in every OECD country surveyed, but national scale-up is rare because of fragmented data, misaligned policies and weak governance. Its main output is a policy checklist of 42 questions to help governments, technologists and health workers move AI from pilots to system-wide use.
Key numbers
- 100% vs 10%OECD countries using AI in health administration vs national-level scale-up of medical imaging AI
- 42questions in the AI in Health Policy Checklist, across 4 pillars and 9 policy categories
- 3%OECD countries developing national legislation specific to AI in health
- 18%OECD countries with a strategy or action plan at the intersection of AI and health
As reported by the source.
English·PDF·opens oecd.org
What it reports
- AI is used in health administration in 100% of OECD countries interviewed; national scale-up for medical imaging applications is 10%.
- Share of countries taking specific actions: AI-and-health strategy or action plan 18%, oversight body 18%, regulatory sandbox for health AI 18%, health technology assessment updated for AI 24%, procurement guidelines updated 11%, national approach to AI in the health workforce 29%, health-specific AI legislation 3%.
- Seven OECD countries have a national strategy addressing AI in healthcare specifically, with several more in development.
- Data foundations are further along: 27 countries have a national health interoperability strategy, 25 have national cloud infrastructure, 20 have established or designated a health data access body.
- 23 OECD countries integrate AI-specific risk classification into medical device regulation; 11 more are working toward AI legislation.
- The checklist has four pillars (enablers, guardrails, engagement, trustworthy AI) and nine categories, each with named leading practices, such as CHAI model cards, the UK Digital and Data Professional Capability Framework and France's citizens' assembly on digital health.
How they got it
Desk research plus the OECD survey on Health Data Governance, the Impact of AI on the Health Workforce and Interoperability, with country results reviewed by member governments. The checklist was built on existing OECD digital health and trustworthy AI frameworks and refined with the OECD AI in Health Expert Group, the Global Digital Health Partnership and the Coalition for Health AI.
Limits
- Country counts describe policies adopted, not clinical or economic outcomes of AI use.
- Survey results were taken at a point in time; the authors note individual responses will change.
- The checklist is deliberately non-prescriptive and does not yet cover agentic AI.
Why it matters for companies
The 100% vs 10% gap matches what many large employers see internally: AI spreads quickly in administration and stalls before scaled, regulated use. The checklist's pillars (data foundations, guidance for scale, workforce capacity, oversight, engagement, trust) transfer directly to any regulated function, including HR and finance. Health providers, insurers and suppliers selling into health systems face a regulatory landscape still being built, so procurement and assurance rules will shift over the next few years.
Summary based on the full text. Read the original for the full argument and data.Open PDF on oecd.org