• 5 min read
If doctors are asked to use AI in their practice, the AMA thinks they shouldn’t be held liable for inevitable mistakes.

Adobe Stock
AI is quickly changing healthcare, as chat tools address medical questions, scribes take notes, and agents answer phone calls. An estimated 75% of 120 polled US health systems have adopted at least one AI tool as of February, healthcare research firm Eliciting Insights reported, up from 59% in 2025.
But these tools are not perfect—and the landscape is quickly becoming litigious. AI has had cases of hallucinations and unconscious bias, and it’s been associated with weakening clinical skills. In response, patients are taking action against hospitals that use AI, concerned the software is breaking privacy laws. Physicians have also filed suits, claiming hospitals are rushing to adopt the technology, risking unnecessary medical errors.
This all leads back to one of contemporary healthcare’s biggest questions: Who should be held responsible when AI makes a mistake? Should it be the companies that create the tech or the doctors and hospitals using it? These questions are becoming more pressing as autonomous AI increasingly enters healthcare.
While some states have established some rules around AI use in medicine, consistent regulations aren’t in place. Organizations like the nonprofit Coalition for Health AI have also popped up to help establish safe guidelines.
“AI can help physicians analyze information, identify patterns, and support clinical decision-making, but patients deserve to know who is responsible for their care,” AMA President Willie Underwood III, said in a release.
A lot of tech companies are relying on a human-in-the-loop method, meaning there’s a clinician who can oversee and fact-check AI output. For example, if an AI scribe takes notes for a doctor, a human doctor follows up to make sure the content’s correct. But doctors can’t catch everything.
Not all cases are black and white, either, and sometimes tools require clearer evidence of safety or efficacy, the AMA wrote in its statements. At the same time, some hospitals are adopting AI tools for their employees, and doctors could end up on the hook for malpractice claims. There haven’t been any cases yet, but it’s likely only a matter of time, Healthcare Dive reported.
The AMA is also pushing for laws that require tech companies to have liability insurance and provide users with clear product limitations and performance data. Contracts with providers, patient disclosures, and consent forms must not be used to free a developer from responsibility for issues with their AI tool, AMA added.
This comes after the AMA released a new framework on Oct. 29 stating that doctors should be able to opt out of using AI tools without penalty if the tech’s “clinical validation, usefulness, or place in the standard of care remains uncertain.”
The AMA declined to provide additional comment beyond its statements.
Jan Carney, president of American College of Physicians (ACP), said the ACP shares similar concerns as the AMA, stating that AI tools should help physicians, not get in their way.
“This patient–physician relationship along with accountability and trust requires the human elements essential to the practice of medicine, elements never reproducible through chatbots,” she told Healthcare Brew in an email.
There is legal precedent for courts holding AI developers responsible for tools and not the end user. To prevent harm, however, systems should have strategies in place to check how tools are working after they are deployed, legal experts told Modern Healthcare.
Sharif Vakili, a primary care physician and CEO and co-founder of telehealth company UpDoc, said via email that doctors already have experience being held accountable for technologies they use, such as an implantable defibrillator that can automatically deliver a shock. When this tech is used, the doctor is responsible for selecting and configuring the device, while the manufacturer must ensure its design and performance.
AI is just another version of that, he said. Though like the AMA, Vakili supports establishing clear rules about who’s accountable for errors stemming from AI output.
“Accountability should follow authority and control. Developers should be responsible for the safety and performance of their systems, health systems for how they deploy and monitor them, and clinicians for the decisions they actually make,” he said.