Only 7 out of 38 OECD countries have a formal artificial intelligence strategy for healthcare, according to a 2026 OECD report on scaling AI in health. This fragmented landscape leaves clinicians unprepared, patients confused, and regulators scrambling to keep pace with technology that's advancing faster than policy can follow.

The numbers paint a sobering picture of unpreparedness. Beyond the lack of formal strategies, just 11% of OECD nations have workforce upskilling programs to train healthcare workers on AI tools, and only 9 countries conduct Health Technology Assessments (HTA) on AI systems. Eight more are in the process of updating their HTA frameworks to evaluate AI's clinical relevance, cost-effectiveness, transparency, and long-term safety.

What makes this gap especially concerning is that there's no internationally comparable way to even measure AI deployment in health systems. Without standardized metrics, countries can't learn from each other's successes or failures. Meanwhile, policymakers face an impossible task: creating regulation at a pace that matches innovation, which has accelerated dramatically with AI.

Why Are Patients and Clinicians at Odds Over AI?

The disconnect between policy and reality is playing out in exam rooms across the world. Healthcare providers report frustration when patients use AI tools like large language models (LLMs), which are AI systems trained on vast amounts of text to generate human-like responses, and then bring questions or insights from those tools to their appointments, treating AI outputs as medical truth. Off the record, some clinicians themselves are adding identifiable patient information into general-purpose LLMs, creating privacy risks that no regulation currently addresses.

Yet patients aren't rejecting AI outright. A 2025 European survey by the European Patients' Forum, translated into multiple languages with nearly 1,000 responses, found that 98% of patients responded positively about AI's potential in healthcare. The dominant concern wasn't skepticism; it was bias and the possibility that AI-driven decisions could be wrong.

"Patients want direct communication from healthcare professionals about AI's role in their care, and where appropriate, AI-specific consent forms," the survey found.

European Patients' Forum, 2025 Patient Attitudes Toward AI in Healthcare Survey

The solution isn't better marketing campaigns. Instead, the actionable insight is that clinicians need to be equipped to talk about AI in patient-appropriate language. When patients discuss AI in the consultation room, that's where trust is built or broken.

What About Young People? Why Are They Missing From AI Governance?

One group is almost entirely absent from AI governance discussions: children and youth. At HIMSS Europe, a major healthcare technology conference, experts emphasized that children are vulnerable to AI systems and AI companions, and can form emotional attachments to them very quickly.

This matters more than it might seem. In many lower- and middle-income countries, 50 to 60% of the population is under 35. Designing AI frameworks without considering how these systems affect young people's health and development isn't just a moral oversight; it's a statistical blind spot.

The pattern is stubbornly consistent, according to research tracking national digital health strategies over more than a decade. Ten years ago, youth health was largely absent from digital health strategies. Today, looking at AI governance frameworks, the same omission appears. When young people do appear, it's through the lens of skills and education, treating them as a future workforce to be trained rather than as current users whose health is already being shaped by these systems.

How Can Countries Strengthen AI Governance in Healthcare?

HealthAI, the Global Agency for Responsible AI in health, recently published recommendations for translating the EU AI Act into action. These five priorities offer a roadmap for countries struggling to build coherent frameworks:

• Optimize AI Supervision: Countries need both horizontal AI supervision (under a dedicated AI authority) and vertical supervision (within the health sector). Joint mock exercises across governmental institutions can stress-test coordination before real deployments expose gaps.
• Address Regulatory Barriers: Health technology assessment and reimbursement pathways are still configured for traditional products. Until that changes, AI scales only for whoever can afford the regulatory entry cost, which excludes most builders and startups.
• Clarify Overlapping Frameworks: The EU AI Act, Medical Device Regulation (MDR), and European Health Data Space all touch AI in health, but different parts of the European Commission need internal coordination, and lessons learned at member-state level need to flow back into framework design.
• Formalize Knowledge Exchange: Instead of duplicating efforts, countries should establish formal mechanisms for sharing what works and what doesn't.
• Engage Multi-Stakeholder Participation: Civil society and patient organizations need to be active actors in legislative and deployment processes, not just tokens at the consultation stage.

The underlying challenge is that accountability historically arrives through litigation rather than foresight. Warnings are already lining up. The Society for Digital Mental Health cautions that general AI models are optimized for conversational fluency and engagement, not designed for clinical accuracy or patient safety. Purpose-built mental health AI, developed using domain-specific clinical data and built with clinical expertise, safety-oriented design, clear boundaries around appropriate use, and crisis detection protocols, is what's needed.

What's Happening in the US While Europe Builds Rules?

Meanwhile, the United States is taking a different approach. In May 2026, the Trump administration deepened its push to position AI as a core component of national security. The Department of Defense announced agreements with eight major technology companies, including Google, OpenAI, Nvidia, Microsoft, Amazon Web Services, SpaceX, Oracle, and startup Reflection, to deploy their AI systems on highly classified military networks for operational use.

The Pentagon framed these deals as advancing its goal to become an "AI-first fighting force." However, more than 600 Google employees signed an open letter urging CEO Sundar Pichai to reject the classified military work, arguing that the broad contract language left few enforceable limits on how the company's AI could be used.

The White House also explored a voluntary federal pre-deployment review process for frontier AI models, which are the most advanced AI systems being developed. The proposed executive order would have required AI companies to provide the government early access to unreleased frontier AI models for up to 90 days before public release to identify security vulnerabilities. However, President Trump postponed signing the order on May 21, citing concerns that it would block AI development. According to reporting, the decision came after prominent tech industry figures, including Elon Musk and Mark Zuckerberg, expressed opposition.

"It would be insane not to give US intelligence agencies early access to AI models," stated Rep. Jim Himes, the ranking Democrat on the House Intelligence Committee.

Rep. Jim Himes, Ranking Democrat, House Intelligence Committee

The contrast between the US and European approaches is striking. Europe is building comprehensive healthcare AI governance frameworks, albeit slowly and incompletely. The US is prioritizing national security and military applications while shelving civilian oversight mechanisms. Neither approach fully addresses the healthcare governance gap that the OECD report highlights.

As AI continues to reshape healthcare, the window for proactive governance is narrowing. Without formal strategies, trained workforces, and clear regulatory pathways, the benefits of AI in healthcare will flow disproportionately to wealthy nations and well-resourced institutions, while risks accumulate everywhere.