Is Not Asking About ChatGPT the Next Malpractice Risk? With Berkeley Law's Vincent Joralemon

Key Podcast Moments

• Vincent explains why a chatbot being wrong may matter less in court than whether the physician asked about it
• Graham and Vincent work through what happens when a hospital administrator mandates an AI tool a physician knows is flawed
• Graham asks whether AI is more like the EHR going down or the CT scanner going down, and Vincent answers with a legal test about outages
• Vincent predicts that most AI liability cases will settle before a court ever decides what the rules are

Most physicians have never been taught the legal framework they practice under every day. Prescribing works the way it does because the law treats us as the filter between a manufacturer and a patient, and that assumption is now being stretched to cover a new kind of tool. It is also being tested by a new kind of patient, one who has already asked a chatbot before walking into your room.

In this episode of How I Doctor, Dr. Graham Walker sits down with Vincent Joralemon, director of the Life Sciences Law and Policy Center at Berkeley Law. Vincent taught high school biology before going to law school at Berkeley and Harvard, and a psychedelics research project pulled him into health law. His recent paper in npj Digital Medicine, “Who bears liability when AI gives bad prescribing advice?”, written with Julia Etkin, asks exactly that question. Graham expected the answer to be technology companies, but Vincent’s argument lands back on physicians.

Vincent explains that there is no standalone body of medical law, only tort liability built from negligence, common law, statutes, and agency rules. Inside that system, the learned intermediary doctrine holds physicians responsible for prescribing decisions, but only to the extent they deviated from the standard of care. He sees two paths for AI: the physician consulting a tool behind the scenes, where the physician still has to judge the advice, and the patient going directly to a chatbot, where a physician may soon need to ask what the patient was told. Because people trust chatbots more than they trusted earlier sources, Vincent says waving the advice away may not satisfy a physician’s duty.

The back half of the conversation turns to what physicians can actually do. Vincent argues that the standard of care is set by other doctors, so documenting why you overrode a recommendation and staying current through conferences, journals, and podcasts is real protection. He and Graham work through de-skilling, using Vincent’s abacus example to show how expected skills shift as tools mature, and what happens when a hospital mandates an AI tool a physician knows is flawed. Vincent closes with predictions on settlements, co-defendant lawsuits, and the biggest mistake a physician can make with AI right now.

This episode is presented in partnership with eNavvi, a clinician-built digital prescribing platform.

ePrescribe anytime anywhere with eNavvi

Top 4 Takeaways

The legal questions around medical AI are not abstract, and they are not being settled in physicians’ favor by default. What Graham and Vincent lay out is where the liability is likely to sit, what a plaintiff’s expert will ask, and what a physician can do today to be defensible tomorrow.