Who’s Liable When AI Makes Medical Decisions?

In the not-so-distant future, a cardiologist might find herself reviewing an echocardiogram flagged by an algorithm she didn't choose, trained on data she has never seen, and deployed by a health system that did not consult her. The AI suggests a diagnosis, but she disagrees and overrides it. The patient recovers, and no one remembers this moment. But if she had acquiesced to the AI's recommendation and the patient suffered harm, she would be the one facing legal scrutiny, with her license on the line.

This scenario underscores a critical issue in the integration of artificial intelligence (AI) into healthcare: liability. As AI systems become more prevalent in clinical decision-making, the question of who is responsible for their outputs becomes increasingly complex and pressing.

Navigating the Liability Landscape

Afnan R. Tariq, co-chair of the SCAI Artificial Intelligence Task Force, and Ami Bhatt, chief innovation officer of the American College of Cardiology and chair of the FDA Digital Health Advisory Committee, highlight a stark reality: in most cases, the physician is held accountable for AI-driven decisions, even when they are based on outputs from algorithms developed by third parties.

"Her name in the chart. Her judgment. Her license. Hers," Tariq and Bhatt write. This means that while engineers, vendors, and health systems play crucial roles in developing and deploying these technologies, the ultimate responsibility often falls on the physician who must interpret and act on AI recommendations.

The implications of this liability structure are significant. Physicians may feel pressured to follow AI suggestions to avoid legal repercussions, even if they believe a different course of action is more appropriate. Conversely, overriding an AI recommendation could lead to defensive medicine, where doctors might hesitate to deviate from algorithmic guidance for fear of being second-guessed in the event of adverse outcomes.

Why It Matters

The stakes are high. As AI continues to reshape clinical decision-making, ensuring that accountability frameworks are fair and effective is essential for patient safety and trust in the healthcare system. Current regulatory and legal structures do not adequately address the unique challenges posed by AI in medicine.

One potential solution is the development of a licensing framework for medical AI systems. This could involve rigorous testing and certification processes to ensure that algorithms meet high standards of accuracy, reliability, and transparency. Clear guidelines on how liability should be shared among developers, vendors, and healthcare providers could help distribute responsibility more equitably.

However, implementing such a framework is no small feat. It would require collaboration between regulatory bodies, technology companies, and healthcare professionals to establish standards that balance innovation with patient safety. The FDA has already taken steps in this direction by establishing the Digital Health Center of Excellence, which aims to promote safe and effective digital health technologies.

Ultimately, the goal should be to create a system where AI enhances rather than hinders clinical decision-making. This means fostering an environment where physicians can confidently use these tools, knowing that they are supported by robust regulatory frameworks and that liability is shared fairly among all stakeholders.

The integration of AI into healthcare offers tremendous potential for improving patient outcomes and streamlining medical processes. But realizing this potential requires addressing the ethical and legal questions surrounding accountability head-on. By working together to develop comprehensive policies and regulations, we can ensure that AI in medicine serves both patients and providers equitably and effectively.