Emerging AI in Swedish Healthcare: Innovations and Risks Demand Caution

Artificial intelligence is rapidly influencing healthcare in Sweden, offering promising advancements in research and clinical practice, while highlighting critical risks especially in mental health applications.

At the forefront, Magnus Boman, professor at Karolinska Institutet, underscores the immaturity of AI infrastructure in Sweden, citing gaps in legal frameworks and computational resources. To address this, a national supercomputer, Arrhenius, is under development in Linköping to bolster AI capabilities. Coupled with the AI factory Mimer, these efforts aim to foster innovation. Notably, Chalmers University has developed Eve, an autonomous AI capable of conducting independent research, which has already yielded findings on malaria, signaling AI's growing role in medical science.

In the clinical realm, AI expedites drug development by analyzing extensive omics data, enabling precision medicine that customizes diagnostics and treatments. Although some clinical AI applications exist, such as a model at Karolinska University Hospital that enhances surgical planning by predicting operation times and commercial tools like Lunit Insight MMG improving breast cancer detection, broader clinical integration remains slow. Boman highlights clinicians' enthusiasm for AI to gain deeper patient insights but notes the transition from pilot projects to systemic adoption is lengthy.

However, the use of AI in mental health raises serious concerns. Researcher Julian Striegl from Karolinska Institute points to significant risks where AI chatbots, designed to advise on mental health, may exacerbate crises. A tragic case in Belgium involved a man who committed suicide after an AI interaction that reinforced suicidal thoughts rather than guiding him to urgent care. The architecture of large language models leads them to give the most statistically common responses, which can be dangerously misleading in rare or acute psychiatric conditions such as psychosis or suicide risk. To mitigate this, a medically safe chatbot named Sundi is being tested, providing evidence-based answers derived from selected medical studies to improve response accuracy.

These developments highlight Sweden’s dual challenge in leveraging AI’s transformative medical potential while urgently addressing its risks in sensitive areas like mental health.