Equitable and Ethical AI Use in Cardiology: Addressing Biases and Privacy Concerns

Artificial intelligence (AI) is a burgeoning technology, developing rapidly across a wide range of industries. Now that it is beginning to make itself known in cardiovascular care, many clinicians have begun to shift their focus from how the technology is implemented to the question of equity and ethical usage.

In a session at the European Society of Cardiology (ESC) Congress 2026 in Munich, Germany, chaired by Sadiya Khan, MD, MSc, the Magerstadt Professor of Cardiovascular Epidemiology, associate professor of medicine, preventive medicine, and medical social sciences, and the director of the center for population science and aging at Northwestern University Feinberg School of Medicine, a series of presentations discussed the ethical and practical pathways for AI implementation in cardiovascular care – as well as addressing the potential roadblocks and pitfalls inherent in this new technology.

“One of the key things that came up in this session – in every single person’s presentation – was that AI has become such a buzzword, and we are all enamored by the opportunity,” Khan told HCPLive in an exclusive interview. “It is of course going to be used, and it is being used, and it should be better implemented because we want to maximize the impact we can have on patient outcomes. The challenge is how we do it, and whether or not it makes things better.”

Although cardiovascular disease accounts for roughly 50% of all mortality in the United States, access to care and advanced therapies are reduced for Black, Indigenous, rural, and lower-income individuals. Additionally, these groups tend to have a higher likelihood of heart failure, stroke, and hypertension. AI’s rapid expansion in recent years has presented the opportunity to bridge these gaps in care access.1

However, AI in itself is also prone to disparities in access and potential inequity. Generative AI, such as the large language models (LLMs) that are increasingly used to collate and interpret medical data, can easily inherit biases and contextual issues from the data used to train them.1

In a review from August 2026, investigators from the University of Pittsburgh and the Cohen Children’s Heart Center in New York analyzed these biases and developed a series of design principles for the creation of equitable generative AI. In particular, they highlighted representation bias, in which case the AI’s training data does not accurately reflect the target population, and deployment bias, when the AI is implemented in a different context from the one for which it was originally designed.1

Additionally, the presence of AI in healthcare is an intrinsically thorny issue in terms of patient privacy. These models are trained on and work with substantial amounts of sensitive data, including imaging, genetic information, and patient history. Without proper security, these data could be vulnerable to unauthorized access or breaches. This issue was a key focus in several of the presentations in Khan’s session, and largely remains an open question for clinicians.2

“Patient privacy is really one of the core things in medicine, and we hold it very dearly – AI threatens the ability of continuing to protect data,” Khan said. “If it is unleashed without guardrails, without clear stopgaps that need to be in place, it has the potential to do that. I think that’s probably one of the key things we need to figure out: who has governance over the data, and how can we maximize their impact in a way that doesn’t threaten patient privacy?”

Editors’ Note: Khan reports no relevant disclosures.

References

1. Kulkarni A, Visweswaran S. Achieving equity in generative artificial intelligence in cardiovascular medicine. JACC: Advances. 2026;5(8):102967. doi:10.1016/j.jacadv.2026.102967

2. Göçer H, Durukan AB. The potential risks of artificial intelligence in cardiovascular care. Turk Gogus Kalp Damar Cerrahisi Derg. 2025;33(2):262-264. Published 2025 Apr 30. doi:10.5606/tgkdc.dergisi.2025.26912