Is AI Liable for Medical Accidents? "Final Clinical Judgment Rests with Physicians"

Attorney Sujin Hur: AI Adoption in Medicine Complicates Physician, Hospital, and Developer Liability
Beware of Exaggerated Claims Such as "99% Cancer Diagnosis" Regarding AI Performance

As artificial intelligence (AI) comes to be widely used in medical settings—from diagnosis and treatment to communication with patients—the legal issues regarding responsibility for medical accidents are also becoming increasingly complex. In particular, even if AI assists with diagnosis, the final responsibility for diagnosis and treatment still lies with the physician, meaning that the use of AI does not necessarily diminish the liability of medical staff, experts point out.

Soojin Hur, Attorney

On August 25, Sujin Hur, Attorney at Law at Bae, Kim & Lee LLC, speaking at the Korean Medical-Bio Journalists Association Academy held at Sangyeonjae in Jung-gu, Seoul, remarked, "The core issue in AI-related medical accidents is not the misdiagnosis itself, but rather how to allocate responsibility."

In the past, medical accidents were solely an issue between the patient and the physician. With the adoption of AI, however, the parties potentially responsible have multiplied to include the hospitals that implement and manage AI systems, as well as the developers involved in disputing defects in the AI itself. Since AI is neither a legal nor a natural person and cannot be the subject of rights and obligations, the scope of legal responsibility is ultimately narrowed down to physicians, hospitals, and developers.

Attorney Hur stated, "AI is merely a tool that assists in diagnosis; the final diagnosis and treatment decisions still rest with the physician." She predicted, "In actual legal disputes, the court will consider whether the physician thoroughly reviewed the AI’s results, comprehensively assessed the patient’s condition, evaluated the need for additional tests, and conducted an independent medical judgment."

Furthermore, the duty of care required from medical professionals may become even more complex with the introduction of AI. This is because, beyond simply reviewing AI-driven results, clinicians must make comprehensive assessments of the patient's condition, decide whether additional tests are needed, and consider relationships between past medical history and current symptoms. Attorney Hur explained, "While the use of AI may make physicians’ work more efficient, there may actually be more aspects that could be deemed as violations of the duty of care if an accident occurs."

The hospital’s responsibility for management also increases. Questions such as whether the hospital sufficiently verified the AI before implementation, whether medical staff were properly trained on how to use it and what to watch out for, whether internal operational guidelines were established, and whether ongoing performance monitoring was conducted after adoption could all become subjects of future legal review.

AI developers are also not free from liability. Since medical AI can be regarded as a ‘product,’ if a defect in the AI itself is the cause of an accident, product liability may become an issue. It will need to be examined whether developers used data biased toward a specific patient group during the design stage, whether sufficient verification was performed before launch, whether performance in real clinical settings was verified, and whether lower performance in certain age or disease groups was concealed despite being known. In particular, if a developer was aware or should have been aware of certain risks yet failed to disclose them, this could result in significant liability in the future.

AI-powered medical advertising has also been identified as a new area of legal risk. Advertising the mere use of AI is not problematic, but once claims exaggerate AI’s performance, medical efficacy, or assert superiority over physicians, this raises issues under medical advertising regulations and labeling/adverting laws.

For instance, expressions that guarantee specific medical outcomes or assert comparison or superiority—such as “AI diagnoses cancer accurately,” “AI is more accurate than doctors,” “AI reduces misdiagnosis,” or “AI finds the best treatment”—are subject to regulation under the Medical Service Act. In contrast, objectively stating how AI is used—such as “We use an AI-based imaging assistance system,” “We provide an AI algorithm-based analysis service,” “We have adopted AI medical devices,” or “We operate a treatment support system utilizing AI”—is relatively safe, according to Attorney Hur.

Attorney Hur warned, "AI can make work at the medical site more convenient but at the same time may introduce new legal risks.” She added, "With the emergence of new correlations of responsibility among medical staff, hospitals, and developers, the number of factors to consider when an accident occurs increases." She also said, "The increased diversity of potentially responsible parties leads to more legal disputes, but this does not necessarily benefit patients. With new issues such as device management responsibility and hospital management systems, conflicts could become more intense for both sides.”

This content was produced with the assistance of AI translation services.