Artificial intelligence is no longer a distant, futuristic concept in medicine. It’s already here, and is becoming an everyday part of clinical practice, supporting diagnosis, clinical decision-making, treatment planning, and patient risk prediction. In fact, according to the World Health Organization, nearly three quarters of EU countries are now using AI-assisted diagnostics in some form, and that number is only expected to grow (Alowais et al., 2023; World Health Organization, 2026).
As the use of AI promises to further revolutionize the sector, what should we know before trusting it with our health?
How AI works
At its core, biomedical AI relies on three main approaches. Machine learning allows systems to learn patterns from data and make predictions without being explicitly programmed for every scenario. Deep learning, a more advanced form of machine learning, uses layered neural networks (loosely inspired by the human brain) to process complex information like medical images or clinical notes. And natural language processing (NLP) enables computers to understand and interpret human language, such as a doctor’s written notes or a patient’s description of their symptoms.
Real examples already in use
In cardiovascular care, AI tools are helping detect arrhythmias from ECG readings and estimating a patient’s individual risk of heart disease. In diabetes management, AI-powered retinal scans can catch signs of diabetic retinopathy earlier than ever before. And in mental health, natural language processing tools are being explored to help identify signs of depression in the way people write or speak, and to support patients between therapy sessions.
So what’s the catch?
Research has shown that AI applications can lead to misinterpretation and the promotion of biases and discrimination (Varona & Suarez, 2022). In essence, AI systems are only as good as the data they learn from. If that data reflects existing inequalities, for example, if it comes mostly from male patients, or predominantly from one ethnic group, the AI will inherit those biases. It will perform well for the people it “knows,” and poorly for everyone else.
Several documented cases have shown AI tools trained on biased datasets producing less accurate results for underrepresented patients, sometimes with real consequences for diagnosis and treatment.
What AEQUITAS is doing about it
This is exactly the challenge our AEQUITAS project, funded by the European Union, was designed to address. By developing training resources, a database documenting real cases of bias in biomedical AI, and a practical regulatory model for hospitals and CSOs, AEQUITAS is helping healthcare professionals and institutions across Europe use these powerful tools more safely, and more fairly, for every patient.
Understanding how AI works, and where it can go wrong, is the first step toward making sure it works for everyone.
References:
Alowais, S.A., Alghamdi, S.S., Alsuhebany, N. et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ 23, 689 (2023). https://doi.org/10.1186/s12909-023-04698-z
Varona, D., & Suárez, J. L. (2022). Discrimination, Bias, Fairness, and Trustworthy AI. Applied Sciences, 12(12), 5826. https://doi.org/10.3390/app12125826
World Health Organization. 2026. “Artificial intelligence is reshaping health systems: state of readiness across the European Union.” World Health Organization.