Professional Growth
What Lawyers Need to Know About AI in Clinical Care
August 03, 2026
At noon on August 6, Walter Sujansky, an adjunct professor of biomedical informatics at Stanford University's School of Medicine, will join Matthew J. Dobson, a trial attorney with the Office of the Attorney General for the District of Columbia, for the hour-long, virtual D.C. Bar CLE class "Uses of Artificial Intelligence in Clinical Medicine."
Geared toward health care attorneys, in-house counsel, risk managers, and professional managers, this course will examine the capabilities and risks of some of the emerging AI tools being deployed in the medical profession. Sujansky and Dobson will also discuss the current federal and state regulatory landscape governing clinical AI.
Here, Sujansky previews the conversation that attendees can look forward to during the program. The interview has been edited for brevity.
Tell us about some of the AI platforms that medical professionals are already using.
The most commonly used type of system is called ambient AI scribe. It is a system that listens. It's a part of the electronic health record or a separate system that listens in the background to the physician–patient conversation during an office visit. It's recording and transcribing everything said during that office visit. Then, using AI logic, it's collating and distilling that into the progress note or the kind of documentation that a clinician would otherwise write manually.
There are also regular large language models or medically specific variants of those that are very commonly being used by clinicians just to do research. Clinicians are [also] using large language models in the context of patient cases to brush up on their memory with regard to the most recent guidelines from professional societies, or the most recent thinking on the diagnosis and treatment of certain illnesses.
For example, if I have a patient who has hyperthyroidism and is also pregnant, I might ask ChatGPT or OpenEvidence, "What are the potential causes of this? What tests should I perform next? What are potential complications I should think about in the context of the patient being pregnant? Are there any modifications to the treatment for this patient because of their current pregnancy?" It will basically give me a tutorial.
If I'm using a tool such as OpenEvidence, it will provide me references for the latest literature, practice guidelines, and articles on that topic if I want to go and check out the sources. More conveniently, it will just give me a summary of that information that specifically addresses the questions I have asked in the context of this particular patient.
Should doctors disclose they are using AI when talking with a patient?
From an ethical point of view, yes, they should. There are different opinions about ethics, but I think most professional societies and ethicists would say, "Yes, that is a requirement because the conversation is being recorded."
From a legal point of view, it depends on what state you are in and what the laws are regarding the recording of conversations in that particular state.
How is patient information safeguarded when medical professionals use ambient AI scribe?
Great question. There are a couple of ways that these recordings and transcriptions are being protected from inappropriate use. There are specific contractual rules. It is important for physicians and physician organizations to be aware of that and to focus on that when they are purchasing and licensing these products to ensure that those protections are in place.
There is an umbrella blanket protection concerning HIPAA privacy rules, which are federal privacy regulations that prohibit health care organizations from sharing patient-specific, patient-identified health information with anyone who does not need to have that information for purposes of treating the patient or arranging payment for that treatment, such as health insurers.
Those rules extend to any agents or business associates of health care providers who are handling patient health information on their behalf. If the recording leaves the premises of the physician's office and goes to the server of the health care technology vendor somewhere else, that server needs to be protected, and that organization needs to prevent that information from being disclosed or shared or hacked. It is their responsibility to do that under federal law.
[Another] way that perhaps one can have some comfort that their information is not floating around is that health care organizations are not retaining these recordings, nor the transcripts. Typically, they're destroying [the transcriptions and recordings] after a couple of weeks; they're deleting them for a variety of reasons, one of which [concerns] patient privacy, but also other legal concerns as far as, just frankly, discoverable evidence in potential litigation.
What are some main takeaways for attorneys who will be attending this course?
The [first] take-home lesson is there are different kinds of AI tools used in clinical medicine, and they are being used for different purposes in different contexts. The kinds of errors that these systems can make, the kinds of reasons for the errors these systems can make, and the implications of the errors that these systems make vary depending on the kind of tool it is and the kind of context in which it is used. It's important to understand all those variances in the course of any kind of legal assessment in use, misuse, or contribution of these tools toward medical errors.
The second is that the regulatory environment is very patchwork, still evolving, imperfect, and subject to political forces. It varies from state to state; it's likely to change within states. It's likely to change federally … Obviously, from a compliance point of view, it will have a direct impact, but also from a point of view of litigation, it will have an impact on the prosecution of civil cases involving clinical AI systems.
The last thing is that, as far as the use and implications of clinical AI in medical malpractice, the case law on that is almost nonexistent at this point. It has all yet to happen. We are at the prenatal stage of the case law for clinical AI in medical malpractice.
Register for "Uses of Artificial Intelligence in Clinical Medicine: An Overview for Legal Professionals."
