. 2021 Apr 6;99(3):629–647. doi: 10.1111/1468-0009.12504

Table 1.

Current Landscape of AI/ML Liability

Type of Liability (Definition)	Implications for Physicians	Implications for Developers or Health Systems
Medical malpractice (Deviating from the standard of care set by the profession)	Physicians may be liable for failing to critically evaluate AI/ML recommendations. This may change as AI/ML systems integrate into clinical care and become the standard of care.	Health systems or practices that employ or credential physicians and other health care practitioners may be liable for practitioners’ errors (“vicarious liability”).
Other negligence (Deviating from the norms set by an industry and courts)	Physicians may be liable for (1) their decision to implement an improper AI/ML system in their practice, or (2) their employees’ negligent treatment decisions related to AI/ML systems (“vicarious liability”).	Hospital liability for negligent credentialing of physicians could extend to failure to properly assess a new AI/ML system. In general, health systems may be liable for failing to provide training, updates, support, maintenance, or equipment for an AI/ML algorithm.
Products liability (Designing a product that caused an injury)	Physicians might be involved in these cases if they work or consult for designers of AI/ML devices.	The law is unsettled in this area. As AI/ML software integrates into care or becomes more complex, algorithm developers may have to contend with liability.

Type of Liability (Definition)

Implications for Physicians

Implications for Developers or Health Systems

Medical malpractice

(Deviating from the standard of care set by the profession)

Physicians may be liable for failing to critically evaluate AI/ML recommendations. This may change as AI/ML systems integrate into clinical care and become the standard of care.

Health systems or practices that employ or credential physicians and other health care practitioners may be liable for practitioners’ errors (“vicarious liability”).

Other negligence

(Deviating from the norms set by an industry and courts)

Physicians may be liable for (1) their decision to implement an improper AI/ML system in their practice, or (2) their employees’ negligent treatment decisions related to AI/ML systems (“vicarious liability”).

Hospital liability for negligent credentialing of physicians could extend to failure to properly assess a new AI/ML system. In general, health systems may be liable for failing to provide training, updates, support, maintenance, or equipment for an AI/ML algorithm.

Products liability

(Designing a product that caused an injury)

Physicians might be involved in these cases if they work or consult for designers of AI/ML devices.

The law is unsettled in this area. As AI/ML software integrates into care or becomes more complex, algorithm developers may have to contend with liability.

Abbreviations: AI, artificial intelligence; ML, machine learning.