Clinical AI may improve care, but it has to meet high standards of evidence and patient safety before it is relied on.
KEY QUESTION
Can AI help clinicians provide better care, and who checks its output?
Priorities are editorial judgements by Cyprus AI Monitor about public value and the attention each sector needs. They are not official strategy rankings, readiness scores or EU AI Act risk categories.
What AI could help with
A clinician-reviewed draft of routine notes could reduce paperwork and leave more time for a consultation.
These are examples of possible uses. They do not describe systems already in use or promise any benefit.
What needs to be in place
Test on the intended patient population. Protect health data, record errors and keep accountable clinical oversight. Do not treat a working digital health portal as evidence that clinical AI is safe.
What the evidence shows
Published figures for Cyprus. These supplementary measures are not part of the index score.
Digital foundationCurated observation
Access to electronic health records
78.75/100
Observation period:
0100
EU average · 2025: 86.51/100
Composite maturity score for people’s technical access to electronic health records.
Limitations
The score describes the access that is technically available. How many patients use the service, how widely clinicians use AI and the quality of treatment are outside its scope.
Sector contextHistorical observation
Medical-sciences research spending
€10.5 million
Observation period:
R&D expenditure classified under medical sciences.
Limitations
Includes research in every field, not only AI, in the government-controlled areas of Cyprus. Spending alone does not show whether research led to successful products or public benefit.
What we still need to know
This release includes no verified Cyprus-wide measure of clinical AI use, patient benefit or error rates.
How progress could be measured
Proposed measures for future data collection. No values have been published for them.
01Clinician time saved after checking and correcting AI output.
02Safety incidents and performance across patient groups in evaluated deployments.
Related source
2026 eHealth indicator study
The 2026 eHealth study measures technical access to records. It explains what its maturity score can and cannot establish.
AI assuranceΔιασφάλιση αξιοπιστίας της ΤΝChecking claims about an AI system’s performance, risks and controls against evidence, before the system is put into use and while it is in use. Review by an independent party adds confidence in the results.Open the glossary ↗High-impact / high-risk AIΤΝ υψηλού αντικτύπου / υψηλού κινδύνουHigh-impact is a general description of uses with significant consequences for people. High-risk is a specific classification under the EU AI Act, based on a system’s intended use and legal criteria, with exceptions.Open the glossary ↗Governed data accessΠρόσβαση σε δεδομένα με κανόνεςRules and controls that set out who may use data, for what purpose, for how long and with which safeguards. Data held by public bodies is not automatically open, nor automatically available for training AI models.Open the glossary ↗