Artificial intelligence systems currently assist in interpreting imaging tests, identifying diseases, triaging inquiries, and directing patients, but the rules meant to protect the public remain far behind. Data presented by the WHO at an international conference in Lisbon shows that only a small minority of countries in the European region have determined who will bear liability when the algorithm makes a mistake, even though its use has already become part of the work routine in many healthcare systems.
The data is based on the most comprehensive survey conducted by the organization to date on the readiness of health systems for artificial intelligence. The survey was conducted in 2024 and 2025 and included responses from 50 out of 53 member states in the organization's European region. According to the findings, 32 countries, which are 64% of respondents, already use tools that assist in diagnosis, mainly in interpreting images and identifying medical findings. Half of the countries have already introduced chatbots designed to provide information, guidance, and support to patients.
However, regulation has remained behind. Only four countries, a mere 8%, reported a dedicated national strategy for integrating artificial intelligence into the healthcare system. Seven additional countries reported that they are still developing such a strategy. At the same low rate of 8%, clear standards for legal liability were also found, defining who bears responsibility in the event that an artificial intelligence system causes an error, harm, or damage to a patient.
Questions of legal and ethical liability
This question is no longer theoretical. A system can mark an X-ray as normal even though a tumor appears in it, estimate that a patient is at low risk even though their condition is deteriorating, or assign low priority to an inquiry that requires rapid treatment. In most cases, the physician is the one who makes the final decision, but as systems become more complex and exert greater influence on the workflow, the question sharpens as to whether responsibility falls on the physician, the hospital, the software developer, the regulatory body that approved the system, or several of them together.
According to the WHO, 86% of countries noted that legal uncertainty is the primary barrier to the safe adoption of the technology. 78% reported that the high cost of systems, infrastructure, and implementation also constitutes an obstacle. The significance is that health systems are required to decide whether to introduce new tools that may streamline work and improve diagnosis, while the legal, ethical, and budgetary frameworks remain partial.
Training of medical staff is also lagging behind the pace of change. Only one in five countries provides AI training to healthcare professionals before they enter the workforce, and only one in four countries offers structured training for workers already in the system. Less than half of the countries have examined whether existing legislation is suitable for the use of new systems, and nearly 40% have not yet formulated ethical guidelines for the use of AI in healthcare.
Bias risks versus real-world benefits
The risk does not stem only from a technical error. AI systems learn from large databases, and if the data on which they were trained does not represent women, minority groups, older adults, or patients with rare diseases, their performance may be less accurate in these populations. An algorithm that appears successful in a laboratory or a specific hospital may act differently when used in another country, in a different population, or under heavy workload conditions.
Alongside the risks, the organization emphasizes that the technology already provides tangible benefits. 98% of countries noted that improving patient care is the primary driver for adopting artificial intelligence. 92% sought to reduce the burden placed on medical staff, and 90% cited the desire to improve efficiency and output. In Portugal, for instance, systems for analyzing medical images help identify chest diseases and bone fractures faster, thereby shortening waiting times in primary care and emergency rooms.
The challenge is to ensure that systems do not become a "black box" that no one understands how it arrived at a recommendation. A physician receiving an alert or diagnosis from a computerized system needs to know its degree of accuracy, in which populations it was tested, when it might fail, and what they should do when the result does not match the patient's condition. Patients also need to know whether a computerized tool was involved in a medical decision concerning them, and what avenues are available to them in the event that damage is caused.
<strong>Privacy issues and the Israeli context</strong><br>
The issue of privacy adds another layer of risk. These systems require vast amounts of medical data, sometimes including diagnoses, medications, genetic tests, mental health information, and documentation of prior illnesses. Uncontrolled use of information could expose patients to leaks, unauthorized commercial use, or decisions based on partial data. The WHO demands that countries set clear rules for data collection, security, secondary use, and the patient's right to know how it is used.
In Israel, too, the debate is particularly relevant. The healthcare system relies on extensive digital medical records, and the Health Ministry operates a National Council for Digital Health and Innovation, whose role is to advise on policy formulation, regulation of medical data, and the integration of new technologies. As the use of systems expands, so does the need to clarify in advance who approves them, who oversees their operation, and who is responsible when they fail.
The Lisbon conference was attended by ministers and senior representatives from 37 countries across all six WHO regions. The organization called on countries to build a national strategy, set liability standards, invest in staff training, and ensure that systems are tested not only in laboratory conditions but also in real-world use. The WHO Regional Director for Europe, Dr. Hans Kluge, warned that the gap between implementation and oversight could become irreversible, especially as patients are already turning to chatbots regarding symptoms before even speaking to a doctor.