WHO Eastern Mediterranean presses for AI readiness before adoption outruns governance
Artificial intelligence arrived in the Eastern Mediterranean Region before the policy scaffolding around it did. In closing remarks in Vienna on 15 October 2026, Dr Adham Ismail argued that the region is building readiness, but not yet at the pace required by the technology now entering health systems, ministries, and clinical workflows.
The message was delivered through history, but the clinical stakes were immediate. Semmelweis, chlorine handwashing, Röntgen’s X-rays, Lister’s antisepsis, and the last naturally acquired smallpox case in Somalia all served the same point: health systems tend to resist innovations until evidence, training, and governance finally catch up. In the case of artificial intelligence, the danger is not resistance alone. It is uneven adoption without the institutions needed to use the tools safely.
Readiness across the region remains fragmented. Twelve of the World Health Organization Eastern Mediterranean Region’s 22 Member States have national digital health strategies or roadmaps. Ten do not. Some countries have already established dedicated structures: Saudi Arabia has created the Saudi Data and AI Authority, Egypt has set up a National Council for AI, and Somalia and Sudan have formed dedicated units for digital health and innovation. Even so, the overall picture remains uneven, and uneven preparedness in health policy almost always creates uneven patient protection.
AI is moving at the speed of the X-ray. Too many health systems are still moving at the speed of Semmelweis.
The comparison is more than rhetorical. X-rays reshaped diagnosis because hospitals quickly built the capacity to use them; antisepsis changed surgery because the method was taught, published, and institutionalized. AI in health will follow the same rule. Algorithms do not improve care simply because they exist. They improve care only when institutions can evaluate them, regulate them, train staff on them, and monitor what they do in the real world. Without that infrastructure, AI can just as easily magnify errors, widen gaps, or entrench unsafe workflows.
Workforce pressure makes the case more urgent. The region has 16.5 nurses and midwives per 10,000 people, less than half the global average of 39.4. It also faces a shortfall of 1.2 million nurses by 2030. In that context, tools that extend the reach of scarce clinicians cannot be treated as optional extras. AI may help triage, documentation, decision support, and administrative burden, but only if health systems first establish clear rules for use, accountability, and oversight.
One sentence captured the practical danger of delay: the technology will not wait for policy.
Cybersecurity was presented as a patient-safety issue rather than a technical footnote. Among health organizations hit by a cyber incident, 72% reported disruption to patient care, and 29% reported increased patient mortality. In a region where health systems already operate under staffing and infrastructure constraints, AI deployment without robust digital protections can turn connectivity into a vulnerability. Unsafe innovation, like an unwashed hand, can kill. The analogy is stark because the consequences are stark: when digital systems fail, clinical care can stall, records can disappear, and patients can be harmed before the problem is fully understood.
The proposed Technical Advisory Group on artificial intelligence was framed as the mechanism that could narrow this gap. A standing TAG-AI would provide Member States with ongoing expert guidance, making AI governance a continuous process rather than a one-off event. That matters because AI policy cannot be a static document drafted once and left untouched while models, data sources, and deployment settings keep changing. Health authorities need repeated review, not a single announcement.
The closing remarks returned repeatedly to trust, education, and governance because those are the three conditions that historically separated good ideas from durable public health gains. Semmelweis waited eleven years to publish his findings. Lister published, explained, and trained, and mortality in his surgical wards fell from 45% to 15%. The lesson is embedded in the history itself: evidence does not spread on its own. Education spreads it. Governance protects it. Investment sustains it.
AI policy in health will face the same test. A national strategy is only a starting point. A regulator is only a starting point. A pilot project is only a starting point. What matters is whether countries move from interest in AI to preparedness for AI, with structures strong enough to evaluate clinical benefit, guard against misuse, and prevent inequity from becoming a design feature.
The speech also drew strength from a local public health success story. In October 1977, in Merca, Somalia, Ali Maow Maalin became the last person in the world to acquire smallpox naturally. He had worked as a vaccinator but never took the vaccine himself because he feared the injection. He later spent the rest of his life vaccinating children against polio and telling hesitant parents not to repeat his mistake. Less than three years later, the World Health Assembly declared smallpox eradicated. Technology, education, and trust converged only after decades of work. The parallel to AI is direct: tools alone do not transform health. Systems do.
The final call was concrete. Twelve countries have strategies today. The goal set before the next meeting was simple: make that 22. WHO signaled readiness to support the journey, but the burden now sits with Member States to close the gap between innovation and preparedness before health systems are forced to catch up under pressure rather than by design.