WHO/Europe met in Lisbon to govern health AI. Three-quarters of the EU already uses AI diagnostics — only three states have a health-AI strategy.

Lisbon just hosted the meeting that health-AI governance in Europe has been waiting for. On 15–16 July 2026, the WHO Regional Office for Europe and the Government of Portugal co-hosted a high-level meeting on AI in health. The framing was blunt: deployment is already here, governance is not.

The numbers from the WHO/Europe readiness report are the story on their own. Three-quarters of the 27 EU member states are already using AI-assisted diagnostics. Two-thirds use chatbots to help patients. But only three states have a health-specific AI strategy, only ten have both named priority areas AND set aside funding, and across the 50 countries in the WHO European region only 20% train staff in AI before they qualify.

That is a governance debt, not a governance system.

What the room actually agreed on

David Novillo Ortiz, Head of Data, AI and Digital Health at WHO/Europe, framed it as a shared problem: how to regulate AI, how to stop it widening inequalities, how to prepare the workforce. The meeting was organised around three pillars — governance, infrastructure, people and institutions — and structured so countries could be listened to, not lectured.

There was consensus on one point that will shape every downstream regulation: clinicians, patients and the public need a seat at the design table. It is not standard practice today.

The regulatory frame that is emerging

Alastair Denniston, Professor of Regulatory Science and Innovation at the University of Birmingham and chair of the UK National Commission into the Regulation of AI in Healthcare, made the argument that regulators are closer to ready than they think. AI is not a new species of thing; it is a therapeutic or a device or a provider, and we already know how to evaluate those on evidence and outcomes.

The harder questions are the ones the industry likes to skip:

• Generalisability. A model trained on one population does not perform the same in another. Post-market surveillance has to catch this.
• Composite systems. When multiple AI tools work in concert across providers, we assess the outcome of the service, not every component — the way we already assess hospitals.
• Liability. Who is accountable when an AI system harms a patient? Product, professional, or provider — the answer changes the incentives.

Why this is a Lisbon story

Portugal is not incidental here. It became the first EU member state to join the HealthAI Global Regulatory Network earlier in 2026, and it hosted the WHO/Europe meeting. That combination — governance appetite plus deployment ground — is why the wellness-and-tech conversation for the next 24 months will keep coming back to this city.

What clinicians, founders and regulators should do next

• Clinicians: demand to be in the design loop for any AI tool your service adopts. The evidence base you would want for a new device is the evidence base you should want here.
• Founders: stop treating regulation as friction. The bar the Commission is describing is the bar that unlocks procurement.
• Regulators: the tools exist. What is missing is a shared post-market surveillance discipline across borders.

TL;DR

• WHO/Europe + Portugal co-hosted a health-AI governance meeting in Lisbon, 15–16 July 2026.
• 75% of EU states use AI diagnostics; 3 have a health-AI strategy; 20% train staff before qualification.
• Three pillars: governance, infrastructure, people & institutions.
• Generalisability, composite systems and liability are the unresolved regulatory questions.
• Portugal's position — first EU state in the HealthAI GRN and this meeting's host — makes Lisbon a fixture for this conversation.

Sources

• Burki T. Governance for artificial intelligence in health. The Lancet World Report, 2026.
• WHO Regional Office for Europe. Artificial intelligence is reshaping health systems: state of readiness across the European Union. Copenhagen: WHO/Europe; 2026. Report