AI in clinical documentation: Who is liable for medical errors?

DR GPT™ perspective

The honest answer is that the law has not settled it. I would rather you learn that here than in a deposition.

The scenario in this piece is deliberately ordinary. An attending opens a chart before rounds. The note is polished, the differential is thorough, a risk model has flagged the patient for deterioration. He did not write most of it. He read it, agreed with most of it, and clicked sign. If that patient decompensates overnight, the question is not only who is responsible. It is who is accountable, and those are different words.

Three parties are plausibly on the hook: the physician who signed, the institution that deployed the tool, and the developer who built it. No court has drawn those lines cleanly yet, and the standard of care is shifting underneath the question while it goes unanswered. Malpractice insurers are moving faster than regulators here, some attaching AI training requirements to coverage and others writing AI-specific exclusions. Worth reading your own policy before it becomes relevant rather than after.

What I tell clinicians to do in the meantime is unglamorous and effective. Document the AI output and your own reasoning as two separate things, in every note. Not the phrase AI-assisted. What the tool produced, what you concluded, and where the two diverged. The physician who records their own reasoning is protected in a way the physician who simply signed is not, and their patients are safer, which is the part that actually matters.

Originally published on KevinMD. Read the full essay at the link below, or see how I build these arguments for a room as a speaker.