South Korea’s medical device surveillance system is facing scrutiny as hundreds of AI-powered medical devices have been approved without a single reported adverse event in over three years. Lawmakers are questioning whether the current oversight mechanisms are adequate for software-based technologies, particularly as AI becomes more integrated into clinical settings.
Data obtained by Rep. Heo Jong-sik revealed that 382 AI-based software medical devices received regulatory approval between January 2023 and March 2026. The number of approvals saw a significant increase, rising from 62 in 2023 to 157 in 2025, marking a more than 150 percent growth in two years. Despite this surge in AI medical device approvals, regulators reported zero adverse events involving these devices from 2023 through June 2026.
This stark contrast to other advanced medical technologies, such as in vitro diagnostic devices, which had 27 reported adverse events during the same period, has raised concerns. Lawmakers suggest that the complete absence of adverse event reports for AI medical devices may indicate a flaw in the reporting system rather than exceptional safety. Heo stated, “The fact that not a single adverse event involving AI medical devices has been reported for more than three years, despite mandatory reporting requirements, could be a sign that the system for detecting software errors is not functioning properly.”
AI medical software assists physicians in interpreting diagnostic information, including CT scans and MRIs, by identifying abnormalities and supporting medical decisions. However, these systems introduce unique risks, such as algorithms failing to detect critical findings or producing inaccurate interpretations. Unlike physical device malfunctions, software errors can be subtle and difficult to trace, potentially leading to delayed treatment or misdiagnosis without immediate indication of the AI’s involvement.
Challenges in AI Medical Device Oversight
The current regulatory framework for medical devices primarily focuses on equipment failures and identifiable malfunctions. AI systems, however, can generate errors that are more challenging to recognize, track, and categorize. Rep. Heo urged regulators to update adverse-event reporting standards to specifically address software-related risks, including algorithmic errors and diagnostic interpretation failures. He also called for an examination of whether hospitals and manufacturers are adhering to their reporting obligations.
As South Korea expands the use of AI in healthcare, regulators face the critical task of ensuring that patient safety monitoring systems evolve in tandem with the rapidly advancing technology. The current situation, with hundreds of approved AI medical devices and no reported adverse events, leaves an open question: does this reflect an outstanding safety record, or are there significant gaps in the reporting and detection system?
The Ministry of Food and Drug Safety data indicates a growing reliance on AI for medical diagnoses, with 55 AI-based software medical devices recorded in the first quarter of 2026 alone. The challenge for South Korean authorities lies in adapting traditional oversight methods to the complexities of AI, ensuring that patient safety remains paramount as this technology becomes a more integral part of medical practice.
