How artificial intelligence could redefine the responsible practice of medicine

Artificial intelligence (AI) didn't just arrive in health care. It exploded. In just two years, health care organizations deployed AI at more than twice the rate of the broader economy, with domain-specific AI tools increasing tenfold over 2023 and another sevenfold over 2024. By comparison, it took 10 years and $27 billion in federal incentives for the health care industry to adopt electronic health records.

From tasks such as dictation and charting to radiologic technology that can scan for early-stage tumors and algorithmic tools that predict sepsis hours before a patient becomes critical, AI is reshaping the clinical landscape at a pace that few anticipated. The promise is extraordinary: faster diagnoses, fewer errors, better outcomes. But alongside AI's promise of innovation comes a growing risk that the medical profession is only beginning to confront.

As adoption accelerates, questions around how to define the standard of care, and who or what must meet it, are looming. Coupled with a trend of "deskilling" in clinical judgment, a new frontier of malpractice liability is threatening to redefine the responsible practice of medicine.

New benchmark for standard of care

In medicine, the "standard of care" typically refers to the level of care, skill, and judgment that a reasonably competent health care provider practicing in the same or similar specialty would exercise under similar circumstances. That standard has long been grounded in clinical judgment: a trained physician, drawing on education and experience, determining the appropriate course of care. AI is reshaping that definition.

Courts, regulators and attorneys are starting to grapple with questions about how AI impacts health care delivery and how it is judged. If widely available technology could have prevented a missed diagnosis, is it negligent not to use it? Or is it negligent to use it improperly?

A provider who once exercised reasonable care without AI may now be measured against a standard that assumes AI-assisted care. What was clinically acceptable five years ago may not satisfy a jury or a medical board today. In this way, AI does not merely improve medicine, but raises the floor, requiring providers to clear a higher bar.

The 'deskilling' dilemma

There is a paradox at the heart of AI-assisted health care: the more reliably an AI tool performs a task, the lower the incentive a clinician has to develop or maintain the skill to perform that task independently.

When a radiologist reviews hundreds of AI-flagged images each day rather than traditional, unreviewed images, the interpretive skills begin to atrophy. When an emergency physician defers to an algorithmic risk score generated by AI rather than synthesizing clinical findings independently and developing a differential diagnosis, pattern recognition fades.

The problem becomes acute the moment AI fails. Biased models trained on one population can perform poorly on another. Software glitches, network outages and integration errors can render a tool unavailable when it is needed most. Providers who have become overly reliant on AI assistance may find themselves unprepared to act independently. The safety net becomes the crutch.

A new frontier of malpractice

An evolving standard of care, combined with concerns about provider deskilling, set the stage for a new generation of malpractice claims.

Although liability is unlikely to shift entirely from individual providers to hospitals that integrate AI tools or the vendors who develop them, it has become increasingly clear that AI will not shield providers from legal liability. It's unlikely courts or juries will be satisfied with an "AI told me so" defense — particularly where the provider failed to exercise independent clinical judgment alongside an AI recommendation.

AI has become deeply embedded in clinical workflows, which necessitates that providers demonstrate not only what the AI recommended, but also how they engaged with that recommendation. Did they corroborate, question or override the output? Why? The clinical record of the future will reflect not only the provider's decision on a plan of care, but also the provider's reasoning in relation to AI's recommended plan of care.

Leveraging AI without surrendering to it

None of this argues against adopting AI in clinical practice. These evolving tools improve patient outcomes, reduce diagnostic delays and support providers managing heavy workloads. The goal is to use AI wisely, as a tool that supports, not replaces, clinical judgment.

It requires a deliberate approach across the care continuum. Medical education should evolve to ensure that trainees develop foundational skills before they learn to rely on AI. Hospitals and health systems should consider policies that define how AI tools are validated, monitored and updated, and that ensure providers understand both the strengths and limitations of AI. Professional societies also have a role in defining what responsible AI-assisted care looks like, so that the standard of care continues to be shaped by clinicians.

AI is a powerful tool that will only continue to dominate health care — the question is not whether to use it, but how providers will do so in a manner consistent with the standard of care.

Olivia Osburn advises corporate health care clients on a wide range of regulatory, compliance and transactional priorities, with an emphasis on mergers, acquisitions, joint ventures, corporate governance and AI. She works closely with physician groups and other health care organizations to advise on operational, structural and legal needs that arise across the health care industry.

Robert Botkin helps clients of all sizes, from Fortune 50 companies to startups, navigate their legal needs tied to privacy, cybersecurity, AI and machine learning. He is known for his ability to translate complex technical concepts and issues into actionable tasks that clients can use to establish effective governance programs in a rapidly changing legal landscape.