“The same AI may be regulated by different entities depending on where it is used, the data that it’s trained on,” said Maya Sandalow.
The Federal Drive with Terry Gerton provides expert insights on current events in the federal community. Read more interviews to keep up with daily news and analysis that affect the federal workforce. Reach out to Terry and the Federal Drive producers with feedback and story ideas at FederalDrive@federalnewsnetwork.com.
Interview transcript
Terry Gerton I’m very happy to have you. The Bipartisan Policy Center’s got an interesting new report out that we’re gonna talk about, Understanding the Regulation of Health AI Tools. When you think about what the average person thinks about when you say health AI tools, I’m thinking about a fitness tracker or a wearable. What are you really talking about here? What does this report cover?
Maya Sandalow Yeah, absolutely. So my organization, the Bipartisan Policy Center has done a lot of work to clarify that exact question. AI is used throughout health care delivery in many different ways. I broadly categorize it into two buckets just to simplify things. And I call the first bucket administrative AI. That’s AI that’s used for kind of operational and administrative purposes. So, as a patient, you might interact with it for scheduling or kind of navigating your benefits. One of the fastest growing administrative AI applications right now is used by healthcare providers to record patient conversations. Those are called AI or ambient scribes. And then health insurers are also using it in claims, something called prior authorization. So they’re the whole slew of operational uses for AI. And then there’s that clinical bucket. There’s traditional medical devices. That’s AI that plays the form of a tool that is used to diagnose or prevent or treat a disease. And then, there’s all these other kinds of tools, such as the rings that people are wearing these days, the Apple watches. So there’s a really wide range of tools. And even within medical devices, a lot of different applications. By far the most common so far is from medical imaging, so we can go into details on kind of all of those as helpful.
Terry Gerton That is a great setup and so what sticks in my mind then is why ask the question of regulation? What is the problem with regulating all of those different buckets of health-related AI?
Earn CPE credit: The latest webinar from the Billington CyberSecurity Cyber and AI Outlook Series will focus on the real-world risks facing AI deployments across the federal landscape. Register now!
Maya Sandalow Yeah, there’s a lot of different regulatory questions depending on the type of AI tool, right? So the Food and Drug Administration, which is an agency within the Health and Human Services, HHS at the federal level, regulates medical devices to make sure that they’re safe and effective before they reach market, right. So the same question applies to AI that’s used in clinical functions. We wanna make sure that they are safe and affective before they’re used on a patient. And then you can imagine there might be a need for policy or oversight, let’s say, for AI that’s used to make decisions about if your health care procedure is going to be covered by health insurance, right? So there’s a lot of different policy questions depending on the type of AI application.
Terry Gerton One of the findings from this report is that many of these tools, as you just described, aren’t classified as medical devices. And so it’s not clear who has regulatory responsibility, or there may be multiple agencies that have regulatory responsibility. What does the playing field look like here?
Maya Sandalow Yeah, it’s a complicated playing field. That’s for sure. The same AI may be regulated by different entities depending on where it is used, the data that it’s trained on. We map out the most common federal agencies that might play a role. So the Center for Medicare and Medicaid Services might play role in coverage and reimbursement decisions related to AI. So, the extent to a healthcare provider is reimbursed for, or a patient is reimbursed for AI used in clinical care. The Office of Civil Rights within HHS oversees something called HIPAA, and that’s big for patient privacy. So any AI used in kind of traditional healthcare delivery and patient data and privacy questions is gonna be overseen by that office. That doesn’t necessarily apply to the rings and the watches that people are wearing that kind of don’t fall within the traditional healthcare delivery system. It’s complicated, but it might not always apply to them. The Federal Trade Commission can get involved if there’s misleading marketing about what these tools are intended to be used for. The Office of the National Coordinator for Health IT, which is a mouthful, is another agency that certified electronic health records. So a lot of patients are familiar and providers use these electronic health record and when certain types of AI are embedded into those tools, then they have purview. So it’s a very complex web, right, that a developer and a provider is kind of navigating depending on the type of AI tool. And that’s just at the federal level. States are also really getting involved, and so the same tool might have strict regulation in one state and not in another.
Terry Gerton Maya Sandalow is associate director for the Bipartisan Policy Center’s health program. Maya, what you’ve just described, I think, goes beyond what most people are imagining. And it sounds like AI is moving faster in healthcare than the regulatory framework to manage it. Everybody’s adopting AI in one way or another, and maybe we don’t have all the rules written yet. What kind of gaps or risks is that creating?
Maya Sandalow It’s a great and important question. I think one central trend to keep in mind, which relates to your question, is that right now, the most common forms of AI used in healthcare are for these administrative purposes. So, things like doctors using it to transcribe notes and insurers using it to review healthcare claims to figure out what the patient provider will be reimbursed for and what will be covered. There’s so much potential in these clinical applications of AI. There’s so much potential for AI to help prevent disease early, diagnose more accurately, treat really complex diseases. And right now, the regulatory framework is so complicated and the payment mechanism for that is not very clear. So there’s kind of a chilling effect of sorts. So, I think there’s a big policy question to be had about how can we create regulatory frameworks and payment policies to make sure that these safe and effective and high value solutions are really reaching the patients who need them most.
Terry Gerton Who’s gonna take the lead in developing that policy framework?
Maya Sandalow Right now, I would say a lot of states are trying to think about this, right? And states are the laboratory for democracy, and so there might be some best practices and lessons learned. But at the same time, these technologies operate at a national level, right, the same developer might be trying to deploy tools throughout the country. And so a lot people think that there’s a need for a federal framework, right. And HHS is the federal agency that oversees healthcare delivery in a lot of ways. And so they’re really actively thinking about that right now.
Sign up for our daily newsletter so you never miss a beat on all things federal
Terry Gerton In the meantime, the opportunities for industry to profit tend to sit in those regulatory gaps, right? So how are the incentives for industry playing into the regulatory process?
Maya Sandalow Yeah, I think this is so important and the goal is to create a regulatory framework that developers right and technology companies are incentivized to go through rather than go around, right? Right now it can be so complicated to get through it and there’s not clear payment on the other side of getting through it that there’s a lot of incentive to say what I am producing, what I developed is not a medical device. It should not be regulated by FDA. There’s a lot of controversy, right? Where maybe a user of an AI tool, a chat bot, for example, might not realize that these aren’t regulated and they’re not intended to be used for diagnosis or treatment purposes. And so we need to create a framework where developers are designed to really go through it, so that patients are interacting with safe and effective tools.
Terry Gerton That safe and effective piece comes up because another function of the regulatory environment is to assign liability when things go bad.
Maya Sandalow Absolutely. That is a big question right now, liability, right? Really, right now a lot of the times providers are held liable for, if something goes wrong, when AI is used in their care. States are thinking about this actively. There’s an important policy development to note in this space. The Office of the National Coordinator for Health IT, ONC, a few years ago put forward this model card or nutrition label policy. Where essentially developers had to give some information that people were comparing to a nutrition label about how their AI functions, right? And it kind of made sure that health care providers were aware of that and shared more of the risk. Early January, the new administration proposed to roll that back. There are proponents and opponents of this proposed rule, but I will say that some providers are really concerned that it will leave them without the tools that they need to make decisions about how AI is used in their health care and potentially increase liability concerns.
Terry Gerton While we’re in this sort of regulatory limbo, what do individuals need to know, both to protect their own privacy and data, but also to get the use of the power of these AI tools that are out there?
Maya Sandalow That’s a fantastic question. I think the most basic answer is don’t assume that an AI solution that you’re interacting with is regulated in a comprehensive way in the same way that the drugs that you might be taking are. And that’s not to say that there aren’t really powerful ways to interact with these solutions, but just be aware that they’re not healthcare providers, right? And honestly, I think it’s also good to talk with your healthcare provider about the tools that you are using. And come up with a plan together about how to really use these in a way that augments care.
Terry Gerton And where do you see federal regulation headed next?
Maya Sandalow Definitely something to keep an eye on. I would say there’s bipartisan agreement about the need for a federal framework because of the state patchwork. The devil is in the details, and that’s where the bipartisan agreement is harder to reach, right? But I would expect that to continue to be a conversation in the years ahead, the need a federal frame work and also to make sure that the agencies at the executive level have the authority and the resources that they would need to play a bigger role in that framework.
Read more: Artificial Intelligence
Copyright © 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
Terry Gerton is host of the Federal Drive and has been working in or with the federal government for more than 40 years.
