Artificial intelligence (AI) in healthcare is often seen as a solution, but its full potential depends on considering the entire clinical context and AI's role within it, including liability. The current AI model in healthcare involves electronic data being processed by an algorithm, which provides a recommendation to a clinician. The clinician can either accept or override the recommendation. However, this model risks making clinicians "liability sinks," where they absorb legal responsibility for errors or adverse outcomes, even when they have limited control over the AI's decisions. This can lead to clinicians acting as a "sense check" on the AI, rather than using their expertise to make decisions. Additionally, clinicians may feel personally responsible for adverse outcomes, leading to mental health issues such as depression, anxiety, and PTSD. In driver assistance systems, a similar issue arises, where the human driver may be held liable for accidents even when the AI is in control. This raises concerns about clinicians being held liable for harmful outcomes from AI-based decision-support systems. The legal framework for AI liability is complex, with current product liability laws not being well-suited to AI. Clinicians and their employers may be held liable for negligence, making them attractive targets for lawsuits. However, this could be addressed by redefining AI as part of the clinical team, rather than a product, which could shift liability to those who employ the AI. Alternative models of AI in healthcare could better integrate patient and clinician input, allowing clinicians to focus on their core role of integrating clinical science with patient context. These models could involve AI providing predictions or data rather than direct recommendations, enabling more meaningful dialogue between clinicians and patients. In conclusion, the current AI model in healthcare risks using clinicians as liability sinks, absorbing responsibility for AI-related errors without sufficient understanding or control. Alternative models that prioritize patient-centered decision-making and integrate AI more effectively could mitigate these risks and ensure that clinicians are not unfairly held liable.Artificial intelligence (AI) in healthcare is often seen as a solution, but its full potential depends on considering the entire clinical context and AI's role within it, including liability. The current AI model in healthcare involves electronic data being processed by an algorithm, which provides a recommendation to a clinician. The clinician can either accept or override the recommendation. However, this model risks making clinicians "liability sinks," where they absorb legal responsibility for errors or adverse outcomes, even when they have limited control over the AI's decisions. This can lead to clinicians acting as a "sense check" on the AI, rather than using their expertise to make decisions. Additionally, clinicians may feel personally responsible for adverse outcomes, leading to mental health issues such as depression, anxiety, and PTSD. In driver assistance systems, a similar issue arises, where the human driver may be held liable for accidents even when the AI is in control. This raises concerns about clinicians being held liable for harmful outcomes from AI-based decision-support systems. The legal framework for AI liability is complex, with current product liability laws not being well-suited to AI. Clinicians and their employers may be held liable for negligence, making them attractive targets for lawsuits. However, this could be addressed by redefining AI as part of the clinical team, rather than a product, which could shift liability to those who employ the AI. Alternative models of AI in healthcare could better integrate patient and clinician input, allowing clinicians to focus on their core role of integrating clinical science with patient context. These models could involve AI providing predictions or data rather than direct recommendations, enabling more meaningful dialogue between clinicians and patients. In conclusion, the current AI model in healthcare risks using clinicians as liability sinks, absorbing responsibility for AI-related errors without sufficient understanding or control. Alternative models that prioritize patient-centered decision-making and integrate AI more effectively could mitigate these risks and ensure that clinicians are not unfairly held liable.