A landmark report from the European Academies Science Advisory Council (EASAC) and the Federation of European Academies of Medicine (FEAM), authored by 22 European scientists, is calling for clear guardrails on AI use across Europe's 29 national healthcare systems. The report acknowledges AI's real benefits — from detecting lesions in radiology to streamlining triage — but insists that high-risk AI systems must remain under strict human oversight, with clinicians retaining final responsibility for patient care.

The authors propose "regulatory sandboxes," controlled testing environments where new AI tools can be evaluated under regulatory supervision before full deployment. They also stress that security, bias mitigation, and robustness checks must be built in from the design stage — not bolted on afterward.

Key Takeaways:

• High-risk AI systems require mandatory human supervision; clinicians must retain final decision-making authority
• "Deskilling" — the gradual erosion of clinical skills when AI routinely handles tasks — is flagged as a patient safety risk
• Equitable access to AI tools and transparency about their performance and limitations are essential policy priorities
• Training should focus on helping clinicians understand AI's capabilities and its limits, not on making them AI experts

Why it matters: As AI becomes embedded in clinical workflows across Europe and beyond, this report sets a framework for responsible adoption — one that keeps patient safety and human judgment at the center of technological progress.