Nearly half of Korean doctors use AI, but legal liability fears slow wider adoption

Some 47.7 percent of doctors in Korea have experience with medical AI, with 83.3 percent using it for image interpretation.

While using medical AI, they expressed the greatest concern about “the risk of medical accidents occurring.”

According to the “2025 Survey on the Use of Medical AI” released by the Korea Health Industry Development Institute, 47.7 percent of Korean physicians have experience using medical AI. (Credit: Getty Images)

On Thursday, the Korea Health Industry Development Institute (KHIDI) released the results of the “2025 Survey on the Use of Medical AI,” conducted by Professor Kim Hun-sung's research team from the Department of Medical Informatics at the Catholic University of Korea.

The survey was conducted to comprehensively understand Korean physicians' utilization experience and awareness levels, as well as the institutional challenges to promoting utilization, amid the expanding use of medical AI technology in clinical settings.

The research team, in cooperation with the Korean Medical Association, conducted an online survey of 2,125 physicians from Oct. 16 to 21, 2025. The results showed that 47.7 percent of respondents reported having experience using medical AI.

Among these, 83.3 percent used it for image interpretation, with the primary purposes being diagnosis (68.0 percent) and screening (51.2 percent). Furthermore, 82.3 percent assessed that AI use improved their work efficiency.

Conversely, “legal risk” was cited as the main reason for not using medical AI.

Even among physicians with AI experience, 69.1 percent cited “unclear legal liability in case of medical accidents” as a reason for hesitation or limited use, and 76.0 percent of those without experience gave the same reason. Regarding liability in accidents, 35.3 percent cited joint responsibility as the most likely scenario, followed by AI developers’ liability (26.9 percent) and individual physicians’ liability (18.0 percent).

In addition, only 5.1 percent of medical institutions had internal guidelines for medical AI use, and only 24.1 percent had provided training. However, 57.5 percent expressed a relatively high willingness to participate in related training in the future.

Furthermore, 69.4 percent of physicians responded that clarifying responsibility and compensation standards is most necessary to promote the use of medical AI, followed by strengthening licensing and certification standards (59.6 percent), data quality management (51.7 percent), and establishing a post-implementation monitoring system (47.9 percent).

Based on the survey results, KHIDI identified three key tasks for the successful introduction of medical AI: securing legal clarity; creating a trust-based ecosystem; and establishing a systematic education system.

“We will support the reflection of the issues and challenges identified in this survey into medical AI policies,” KHIDI said. “Considering the evolving AI technology and changing medical landscape, we will continue to secure in-depth and objective policy grounds through follow-up surveys.”