AI in Healthcare — From the Exam Room to the Courtroom

Introduction (00:01):
Welcome to Health Law Talk presented by Chehardy Sherman Williams. Health law broken down through expert discussion, real client issues, and real life experiences. Breaking barriers to understanding complex healthcare issues is our job.

Conrad Meyer (00:21):
And good morning, good night, good afternoon. Whatever time is you’re listening to this, welcome to another episode of Health Law Talk here at Chehardy Sherman Williams. And today George Mueller is with me. Happy to have him with us, George.

George Mueller (00:36):
Good morning, Conrad. How you doing?

Conrad Meyer (00:38):
Good to see you. 6.8 billion

George Mueller (00:43):
With a

Conrad Meyer (00:44):
B. Is what I’m hearing today and what we’re going to be discussing today, that is exactly how much money the DOJ recovered in false claims last year, which is the most in the history of the statute. And what keeps up every healthcare executive at night that they just decided or just announced that AI driven manipulation of electronic healthcare record is now a top enforcement priority. So if your AI vendor has come to you recently pitching the latest and greatest efficiency booster, it might be your biggest compliance liability.

George Mueller (01:21):
Yeah. I mean, a lot of EHRs have AI baked into them and a number of other things. AI has become increasingly a significant and very helpful and useful tool. Certainly in many instances takes a lot of time out of voluminous tasks and can compress a lot of data, process a lot of data very accurately. I think what you’re seeing here is you have either kind of upbilling momentum or you have diagnostic assistance and trends are starting to form over large quantities of billing data that suggests that it starts to, I guess, shape up to look like what the Feds would call potential False Claims Act issues. But I know there’s a lot out going. I mean, in some respects there’s a lot of AI ability to detect fraud and in some ways with volumes of billing being assisted by it, you’re going to have instances where maybe billing trends or I call it upward billing momentum gets cited as something that’s a claim.

Conrad Meyer (02:28):
And that’s what we’re going to cover today. We’re going to go over the breakdown of the record breaking FCA enforcement numbers. We’re going to hit the DOJ focus on AI-driven fraud, what that means for you or your practice or your facility. And then we’re going to walk through some of the new state AI regulations, specifically the Texas Trager Act, California’s AB489 and some few others and what they require. And then we’re finally going to tie everything together with practical steps, what every healthcare organization should be taking right now to do to protect and minimize your exposure in this era.

George Mueller (03:00):
Yeah. I know as you mentioned, both in Texas and in California and in Colorado, you’ve got a bunch of new laws that have taken effect in a number of states and then you’ve got the DOJ also kind of separately pushing up on its enforcement targeting AI and healthcare fraud. And so I guess for us, we have primarily a lot of Louisiana clients in the Gulf South

Conrad Meyer (03:22):
Who

George Mueller (03:23):
Use these tools and they probably just because of their novelty aren’t even aware of what their compliance obligations are. So whether you’re a physician and you just have that EHR with AI assisted diagnostic tools, your hospital administrator looking at new technology and there’s plenty of it out there or your practice manager trying to figure out your compliance obligations, I think we’ve got a lot to cover for you today.

Conrad Meyer (03:49):
So we’re going to unpack it. We’re going to start with the $6.8 billion wake up call, the FCA enforcement and AI fraud. So January 12th, the DOJ announced that the FCA, the False Claims Act recoveries hit $6.8 billion for last year, which was a record. This was a record, George.

George Mueller (04:07):
That’s a staggering

Conrad Meyer (04:08):
Number. It’s a staggering number. And we see it all the time with these fraud and abuse CLEs and the discussion about it. So it is 100% a target for DOJ.

George Mueller (04:19):
Yeah. I think you were looking at some of the materials for this call today. I think the number of new government initiated healthcare matters hit a record to 183, which is up from just 87 from the prior year. Wow,

Conrad Meyer (04:31):
That’s the double.

George Mueller (04:32):
My math,

Conrad Meyer (04:33):
It’s like it’s

George Mueller (04:33):
More than double,

Conrad Meyer (04:34):
Right?

George Mueller (04:35):
Yeah. And so it’s not just more money, but it’s more investigations. And candidly, they’re probably being aided by AI.

Conrad Meyer (04:42):
I think so.

George Mueller (04:43):
Because I don’t think they hired that many more employees in this enforcement area. So I think it’s fair to say that AI is going to be used to create some of these problems and AI is going to be used to help remedy some of these problems.

Conrad Meyer (04:54):
And I think that’s important because every year what DOJ does, they announce their priorities. And so at the top of the list of this year, they’re calling AI driven EHRs manipulation top priority, top priority.

George Mueller (05:06):
Yeah, no. And so what we led with at the beginning of the talk is you’ve got these EHRs that all practices practically have to have them now, of course. They generate these automated prompts, they have default selections or they have nudges or other things that assist in diagnostic criteria, treatment suggestions, and then they feed into the claim submissions. And you have to watch where in some cases something may not be medically necessary, air quotes, or in a worst case, services that were never actually provided or in some other cases, I know a lot of payers see this where you have actual notes from a physician saying X, but then the nudge says Y and for whatever reason, maybe the billing represents Y and not X.

Conrad Meyer (05:51):
They don’t come aside.view

George Mueller (05:52):
Of recoupment. They’re like, “Wait, your own physician said this was X, but you billed for Y, what’s happening here?” And I don’t think they’re intentional. I think you’ve just got, it’s volume driven and it’s erroneous and because of the novelty of the technology, people aren’t attuned to catching it yet, which is why we’re talking

Conrad Meyer (06:10):
About it. And we haven’t got into it, but we have the 60 day overpayment rule and when you see that and what do you flag? So I mean, if you’re looking at your E&M codes to your point and you’re traditionally at a 24%, maybe a level five, a 20% level four and like a 42% level three, but then suddenly after you- A lot of your

George Mueller (06:30):
Threes

Conrad Meyer (06:31):
Turn into four. Now you got a level five at like 35, 40%. You might want to look at that a little more closely.

George Mueller (06:37):
Yeah. We’ve seen that organically in situations that weren’t aided by AI and so I’m sure it’s going to leap out even more when A, it’s being caused by AI. And then obviously I think the feds are using AI to review it.

Conrad Meyer (06:51):
So

George Mueller (06:51):
It’s going to get

Conrad Meyer (06:52):
Caught. And I agree with you. I don’t think it’s purposeful. I mean, I think what people don’t realize is your EHR vendor, the one that pitched you on this might have integrated these AI features to auto-populate some E&M codes and maybe the E&M levels maybe suggest additional procedures, maybe have some notices that pre-fill the documentation for, I guess revenue cycle management they call it in the industry. And each one of those suggestions, if it results in a claim that’s being submitted to Medicare or Medicaid that is not supported by the actual, to your point, the actual encounter, that is a potential false claim right there.

George Mueller (07:28):
Yeah. And you wonder if the data sets that are being used to generate the three to the four, the four to the five and the other nudges, if that’s just static data or if it’s learning and if it’s learning from its own mistakes, it could compound. I’m not a code head. I don’t know that, but you can imagine that I think A, the awareness of it and B, how to adjust it and adjust down from it to minimize what we would call erroneous billing episodes has got to be adjusted. And that’s got to be something with the EHR provider and then the responsible party, the practice, hospital, physician, et cetera.

Conrad Meyer (07:59):
But you know it bites. I mean, if you get hit with an FCA, it bites. No,

George Mueller (08:03):
So if you look at, I think it’s what, section 3700, 3729, I think you get treble damages, right? And that’s the big monster that everyone loves on this if you’re recovering.

Conrad Meyer (08:14):
And that’s per claim.

George Mueller (08:15):
Yeah. You get civil penalties, you get like 27,000 per false claim that hits fast on a high volume.

Conrad Meyer (08:23):
Absolutely. I know you deal with that because you do a lot of M&A and you do a lot of due diligence and so I’m sure the implications when you’re doing an M&A deal hit you there as well.

George Mueller (08:34):
It kind of reminds when cybersecurity became a big thing, it’s become the new big issue in our transaction documents, right? We’ve seen the AI reps and warranties and compliance and indemnities related to it. It’s not just a line or two,

Introduction (08:52):
It’s

George Mueller (08:52):
Significant. And of course, I think in healthcare transactions, we’ve been telling people on both sides of it to conduct specific AI compliance and due diligence, right? So if you’re a seller, you kind of want to know what you’re supposed to be. If you’re a buyer, if you’re acquiring a practice or a healthcare company, you need to understand what AI tools are embedded in the target, like in their workflows, what kind of claims tools they have and anywhere else that AI is being used. And this could be not just embedded in the tools, but you may have individuals who have AI tools that they’re using kind of on their own to support one of their workflows. And so you’ve got to identify all those because success reliability is a real risk and you could, even if you come in with new software or updated software, the retrospective can be obviously significant potential exposure.

Conrad Meyer (09:45):
And to piggyback off of that, we saw a notable development last year where DOJ pursued a false claims claim against a PE firm, a private equity firm. So not just the portfolio company, but the actual private equity fund is- Well, their hands are probably

George Mueller (09:59):
Handling the

Conrad Meyer (10:00):
Billing, right?

George Mueller (10:00):
They’re probably integrated

Conrad Meyer (10:01):
Into it. So that’s true because the sponsor’s personnel have been actively involved in managing the IT. They’ve been managing operations as most PE portfolios they want to do. So if you’re a private equity sponsor and you have a healthcare portfolio, you better be watching this.

George Mueller (10:16):
Yeah. And you know who is watching it is the rep and warrant insurance providers.

Conrad Meyer (10:19):
Yes.

George Mueller (10:20):
They’re usually a good, I don’t want to call them a canary in the coal mine because all those canaries are still living, but they’re very actively upgrading, focusing, adding emphasis to the specific risk in their healthcare deals. It’s almost like a separate little line item that they’re giving a lot more focus to. But I’ll tell you this, you talk about that, the peak thing. Let’s say, what if you’re like a hospital CFO, right? Yeah. Everyone, they come, they do the dog and pony show. We’ve got the best EHR, it’s the latest and greatest. We’re going to take the place of billing clerks and we’re going to do it faster. And by the way, you’re going to get more money and we’re all that and because it’s got predictive coding, you got automated documentation assistance, clinical decision support. I really like that term, right? Those are nudges, but your coding productivity goes up,

Conrad Meyer (11:11):
Your

George Mueller (11:12):
Revenue goes up, then who comes knocking, right? Your government payer. That’s right. They get a civil investigative demand. US attorney wants to say, “Hey, I want to see everything you did for the last three years that’s been influenced by AI.” And you’re like, “Well, gosh, we’ll be at AI for the past year. Oh, great. We’ll get a good break point on your data.” And then you can immediately, as you were saying earlier, what ends up happening is the trends become very stark and obvious. And so they need to be supported, right? And kind of getting correlation or alignment with actual diagnostic information versus what the computers are generating, that’s the daylight and the delta that you want to minimize so that these audit things don’t become enforcement actions.

Conrad Meyer (11:53):
And we can’t ignore the whistleblower action, the QTAMs, because they are the major driver of FCA enforcement. And for those of you who don’t know what that is, a qui tam whistleblower is someone who comes and brings the claim. Sometimes there’s some folks who are embedded in a company that just know that the behavior or the false claims are going on and they come and they whistleblow and they get a cut. They get a piece. If the government intervenes and they win and they settle, those whistleblowers get a piece. So we’re already seeing where coders and compliance officers are coming forward with quytam suits alleging that AI generated coding systems suggestions led to a systematic upcoding. I’ve seen that. I’ve had that in one of … And it wasn’t really the AI that was doing it, George. It was a new vendor that suddenly … And it was all urgent care.

(12:41):
It was an urgent care company that basically had their E&M codes on level four, level threes and level fours and level fives. Suddenly level fours and level fives doubled, doubled in a year, which is just historically over five years never happened.

George Mueller (12:57):
Either the people in the geography got a lot more sick or there

Conrad Meyer (13:00):
Was

George Mueller (13:00):
Just some billing momentum in a direction that may not have been supported.

Conrad Meyer (13:05):
Correct.

George Mueller (13:05):
Hopefully they got all that straight.

Conrad Meyer (13:07):
They did. They did. But basically you see a lot of … You can’t ignore the QTAM whistleblower because those people get a recovery, a portion of the recovery. So that’s a very powerful incentive.

George Mueller (13:16):
And so I guess one of the questions that comes to mind from a transactional standpoint is who bears responsibility for something like that when you’ve got an AI tool generating an inaccurate claim? Is it the provider who signed off for it, the practice that bought and is managing the tool, the vendor that built it, the feds, they’d probably say everybody and y’all figure it out. We’re getting our money

Conrad Meyer (13:38):
Back. And to me, it’s real simple. From a regulatory standpoint, the short answer is under False claims and even the CMS 1500, even though it’s electronic says that the provider who submits the claim is on the hook, period, end of story. They don’t care if you intended to defraud, they don’t care who is it, you’re on the hook.

George Mueller (14:02):
Yeah. And then I think for our practitioners and physicians, other licens providers that we represent, obviously what you want to try to do when you acquire and use and license these tools or when you acquire companies that are using these,

(14:18):
Your indemnities and your contractual review is critical to try to get backstops for those things. It’s because one of these scenarios where it’s not my fault, but it’s certainly my problem. And so what we try to recommend, what we see is that to the extent you can and you’re the physician, you got to get some indemnity somewhere, right to the extent you can grab them. Certainly we do it in our deal work to the extent we’re a buyer and if you’re a seller, you want to try to have as clean a bill of health as possible and it’s kind of a reconcilable set of AI data for your billing. Well,

Conrad Meyer (14:53):
I’ll tell you this. At one point that our listeners need to know that the False Claims Act does not require that you intend to defraud the government. It requires that you acted with reckless disregard for the truth of the claim, reckless disregard. So if you’re blindly accepting what AI does without any kind of- liability away, right? You can’t just say, okay, I’ve got 10 subagents running billing audits and I’m just sitting by drinking my coffee. If you’re not doing human review, that would easily be a provable point you’re just reckless disregard.

George Mueller (15:23):
You’re letting it run on its own. And I think that Supreme Court decision was a shoot versus a super value, the knowledge and the specific intent of the Center standard, I think got even more provider unfriendly. The court in that case said, think if you subjectively believe or have reason to believe that your claims are false, that satisfies the False Claims Act’s knowledge requirement. Even if your interpretation of the rules is objectively

Introduction (15:50):
Reasonable.

George Mueller (15:50):
So you’re going to have a very enforcement friendly court probably looking at whether you subjectively had a reason to believe your claims are false. Now that I’d be very concerned about that because I think it’s going to be pretty easy to prove

Conrad Meyer (16:06):
For the

George Mueller (16:06):
Government.

Conrad Meyer (16:07):
And I think that’s why the DOJ and Health and Human Services created this False Claims Act working group that was formalized last year. So it’s not DOJ acting alone anymore. They’re combining data with HHS and sharing strategies and coordinating. I think we’re going to see in 2026 a tremendous amount of not only AI enabled and coding issues, I think you’re going to see it’s expand a lot more on AI issues in 2026. We’re going to have to see how that plays out.

George Mueller (16:37):
As a news blurb, you just reminded me, did they say that 18% of all home health service claims in the United States are coming from one

Conrad Meyer (16:46):
County

George Mueller (16:47):
State of California?

Conrad Meyer (16:48):
Right. How is that not fraud?

George Mueller (16:51):
Well, I know California thinks the most populous state in the Union, is that correct?

Conrad Meyer (16:57):
But 20%- That’s staggering. 18%.

George Mueller (17:02):
That’s a

Conrad Meyer (17:02):
Lot. Well, that’s 18, but still. So almost one fifth of the countries, right?

George Mueller (17:08):
Entire

Conrad Meyer (17:09):
Home health. That is a lot.

George Mueller (17:11):
That might be suggestive of controversy and I bet AI is looking into it.

Conrad Meyer (17:16):
I would be surprised if it wouldn’t be. It’s

George Mueller (17:18):
Probably how they discovered it.

Conrad Meyer (17:19):
All right. So now let’s hit another topic, the new AI regulatory patchwork, what do providers, what should they know? So let’s talk about on the regulatory side. So George, the fundamental problem, Congress has not passed AI legislation yet. And I think- I

George Mueller (17:35):
Don’t know if that’s a problem. I

Conrad Meyer (17:38):
Don’t know. They don’t need to,

George Mueller (17:40):
Right? But look, it-

Conrad Meyer (17:42):
They don’t know. It’s like the wild west, right? And look at the people though passing the legislation. I mean, I guarantee you half those or more than half have no idea what ChatGPT, Claude, perplexity, they don’t know any of that stuff. They don’t know what it does.

George Mueller (17:58):
It’s interesting because we’re kind of in the vacuum of knowledge. We’ve got this extraordinary increase acceleration of technology and its availability. It’s definitely outstripped regulation for it. You get a little concerned that if they’ve regulated too quickly, it’s going to be vast, maybe overreaching, and then what happens with that? It takes probably some guarded analysis to see what, if anything, they need to do with that.

Conrad Meyer (18:27):
Well, it’s interesting because in typical fashion, the states have taken the bull by the horns, literally.

George Mueller (18:33):
And you’re

Conrad Meyer (18:34):
Going

George Mueller (18:35):
To get a tower back.

Conrad Meyer (18:36):
And we’re going to go in Texas. We’re talking about Texas and what they’ve done in California, what they’ve done, but the states have not waited and they have decided to tackle AI directly as it affects healthcare providers. Sure.

George Mueller (18:46):
In fact, I think that the two I was looking at in connection with this talk was, we don’t have one in Louisiana, but if next door in Texas or California, they pass laws. They’ve got some new laws there.

Conrad Meyer (18:58):
They do. And let’s start with Texas. We’re going to start there. So arguably the most far reaching legislation Texas passed what’s called the Responsible Artificial Intelligence Governance Act, Trega, which took eff this year. So it took effect last month, probably one of the broadest-grill.

George Mueller (19:15):
Is that a grill name?

Conrad Meyer (19:16):
It’s Traeger.That’s Trager.

George Mueller (19:19):
Trying to give some common

Conrad Meyer (19:21):
Relief for our listeners. No, that’s fine. We’ll let everybody know we do not get any kind of promotional gift from Traeger, right? I think a

George Mueller (19:26):
Lot of people say Traeger though, when they mean Trager.

Conrad Meyer (19:28):
They could. Absolutely. In New Orleans, absolutely. Absolutely. I don’t know if they do that in Texas. They have traditional stickers. Ours

George Mueller (19:36):
Longer in Texas, but it’s not existed in Louisiana.

Conrad Meyer (19:38):
But this is one of the most broadest AI laws in the country. So George, can you walk us through a little bit about Trega? What does it entail?

George Mueller (19:46):
Trega, not the wood pellet burning grill. Look, this thing establishes some broad governance requirements for what it calls, air quotes you can’t see,

Introduction (19:57):
High

George Mueller (19:57):
Risk AI systems. And so this is not healthcare specific, but healthcare AI is explicitly included in what’s covered by this law. And so what happens is for licensed healthcare practitioners, physicians, nurses, therapists, this law requires a conspicuous written warning or disclosure to patients whenever AI is used in a diagnosis of treatment. So think about similar to the type of boilerplate open and obvious disclosure if say a physician owns an interest in the ASC where surgery’s

Conrad Meyer (20:33):
Going to be done or a diagnostic facility

George Mueller (20:35):
And you have that-

Conrad Meyer (20:35):
They got to tell them.

George Mueller (20:36):
Disclosure and think you got to let it’s got to be open and obvious and then they’ve got to do it to where it’s not buried inside of a

Conrad Meyer (20:42):
Document. So let me be very specific about what that means from a healthcare regulatory standpoint. So if a physician in Texas uses an AI tool to help, for example, interpret a radiology image or uses an AI tool for some sort of clinical decision support tool that maybe helps with a differential diagnosis that incorporates machine learning or uses some AI diagnostic platform, they must tell the patient in a written disclosure before the encounter or at the time of service. So they have to give notice.

George Mueller (21:13):
Yeah. And it’s got to be, as I said before, conspicuous, clear, prominent, language patient can understand. It can’t be gobbledygook or buried somewhere in a

Conrad Meyer (21:22):
20-stage engine. Some minuscule font. That’s

George Mueller (21:24):
Right.

Conrad Meyer (21:25):
So now look, let’s hear the Louisiana angle because most of our listeners here are in Louisiana. Now Louisiana, again, in fine fashion, has not caught up and has not done this. I was talking to Rory about this last week about whether or not the state board of medical examiners was going … They have to be looking at this. They’ve got to be.

George Mueller (21:46):
Anytime that a legislative body or a state hasn’t passed something yet, I praise them for their either reluctance or their slowness to move because every time you pass laws you get-

Conrad Meyer (22:01):
No, I get it. It’s kind of like you have to wait for version two of something, right? You got to wait it out. I think they

George Mueller (22:05):
Need to kind of figure out what actually need … So again, we have this wash of technology that’s come across us.

Conrad Meyer (22:12):
And it’s moving at warp speed.

George Mueller (22:14):
Very fast. But what actually would need to be fixed, right? I mean, before they go doing something and putting severe limits on the use of the technology

Conrad Meyer (22:22):
I don’t know if that the limits is, but I like the disclosure part of Texas law, but we’ll see because I don’t know. I don’t know what I don’t know.

George Mueller (22:28):
Yeah. I think this is here to some degree to let people know it’s kind of good government stuff. I don’t know if it’s-

Conrad Meyer (22:34):
But what I think our Louisiana-

George Mueller (22:36):
Let’s say you’re going to this place and they use

Conrad Meyer (22:38):
AI,

George Mueller (22:38):
Which a lot of people do, especially in the diagnostic things like radiology and whatnot. So because in radiology, geez, if you have the data set for how many reads, for how many specific type of diagnoses to correlate those, that’s an extraordinary tool, right? That magnifies the ability to get correlation and diagnosis and good medical outcomes and to reduce errors. That I think is a good thing to have. If you’re going to go get a read and someone says, “Are you going to leave the place because they use AI?” Probably not.

Conrad Meyer (23:12):
No.

George Mueller (23:13):
Thank

Conrad Meyer (23:13):
God

George Mueller (23:14):
They’re in the 21st century, right? So it’s good, but I think time will out what if anything needs to be fixed with this, but-

Conrad Meyer (23:23):
But wait, I think we need to make sure our Louisiana providers know that if you’re a Louisiana provider and you are licensed in Texas and you are providing services in Texas- Oh

George Mueller (23:33):
Look, if you’re-

Conrad Meyer (23:33):
You better be compliant with Trega.

George Mueller (23:35):
If you’re over on the Sabian River or you’ve got telehealth in multiple states, then I think from a business formation standpoint, what we’re telling people is if you’re expanding into Texas or you’re projecting into Texas because you’re multi-state licensee or because you otherwise are close enough to Texas, you’ve got to build this into your operations from day one, right? I mean, you have to be compliant with it because invariably if you have a Texas licensee, you’re covered by Texas, you’d be covered by this.

Conrad Meyer (24:07):
I agree. Now let’s move now, that takes care of Texas. Let’s go to California. So California passed AB49, which is also effective January of this year, it takes a much different approach. So rather than requiring a disclosure or notice to patients, what this bill does is prohibits AI developers and deployers from using terms, phrases, or design elements that indicate or imply that the AI system possesses a healthcare license or certification. That’s interesting.

George Mueller (24:36):
Right. Yeah. No, it’s a slightly different version of some good government taking place here. To me, it seemed like the doctor chatbot law, right? So if you have one of these AI chatbots that presents or says, “Hey, I’m your doctor,” or uses an MD badge or other Indicia of being an actual provider, that is now illegal in California, right? So if you have an AI system that presents itself as a licensed therapist or uses a stethoscope icon in a clinical context, that’s prohibited, right So basically you can’t brand an AI interface as an actual healthcare provider, which you would think is somewhat obvious, but if you think about it, and I think there’s some validity here, when you go to do a chat for anything, like if you have live chat about a product you’re buying or trying to facilitate a return, you’re trying to figure something out if you’re talking to someone who says, “I’m your doctor,” you may think you’re chatting with a doctor.

(25:33):
And so if you’re not chatting with a doctor, then I think it’s good that people who are accessing that actually are know that are reminded of that. I think that’s because that-

Conrad Meyer (25:43):
I think the common person who may not understand that would really believe they’re chatting with a physician.

George Mueller (25:48):
That’s right. And someone that doesn’t frequently accesses that or is just accessing

Conrad Meyer (25:51):
It personally. They would have no idea.

George Mueller (25:54):
Depending on a multifactor analysis of who they are and what they’re about, I think it’s important. I think it’s helpful,

Conrad Meyer (26:01):
Right? Well, California didn’t stop there. So California also enacted SB 243. They call this the companion chatbots. And I see people use this a lot where the AI systems that engage in these extended conversations or interactions often used for mental health or wellness or things like that. Now under the California SB 243, these systems must notify users that they’re interacting with AI. So they must have protocols to handle the discussions, especially with suicidal ideations.

George Mueller (26:32):
Yeah. Again, I think it’s important for people to know who they’re talking to have a natural tendency to kind of humanize and personify these chatbots. So I think it’s good necessary.

Conrad Meyer (26:44):
It seems

George Mueller (26:44):
Kind of intuitive, but-

Conrad Meyer (26:46):
Common sensical, but … Yeah,

George Mueller (26:47):
Exactly right. So there’s nothing common about common sense though, I’ll tell you that. So it does get interesting as even healthcare organizations, if they’re deploying these kind of patient facing AI, if you’ve got one of these patient portal chatbots, symptom checker or any other AI intake tool that interacts directly with your patients, you need to evaluate your compliance obligations in every state where you’re located.

Conrad Meyer (27:10):
I agree. I agree. And you know what? It’s so easy now, George, to make an app. It is so easy to code and draw and make an app webpages now with these AI tools. So yeah, I agree. It goes back to telehealth when we had to go to licensure requirements by state by state. So if you’re not doing that, you’re going to be in trouble. Now let’s switch to Colorado because we’re going with a theme now of the states that have sort of grabbed the bull by the horns and have done their own AI law. So California passed SB 205, which was originally set to take effect in February, but was delayed now. It’s delayed till June and this law focuses on preventing algorithmic discrimination, essentially making sure that your AI system in healthcare doesn’t produce biased outcomes based on race, gender, and other protected characteristics.

George Mueller (28:01):
I think this is an interesting law because for the practitioners, physicians we’ve worked with over the decades, we know that different people have different genetic predispositions to certain disease processes.

Conrad Meyer (28:12):
But you also know, I don’t know, I’m kind of going off here, but remember back when Gemini first came out or forgot if it was Gemini or something else and people were asking about questions of Gemini about historical truth of the United States and it came up with this crazy like really biased response. So I think that’s what they’re trying to avoid.

George Mueller (28:30):
Yeah. And so my point was I think the law versus the actual practice of medicine has to really thread the needle so far as knowing your patient, knowing based on who they are, what their dispositions are for purposes of good diagnostic medicine.

Conrad Meyer (28:46):
Correct.

George Mueller (28:47):
I mean, I’m a melanoma survivor from where my descendants are from, so to speak, descendants, like 60% more likely to get melanoma because of where in Europe my ancestors are from. So that’s an important factor for my dermatologist to know in order to appreciate risk. But I think this is important because what you don’t want to have is what I just mentioned is one thing, but when you’ve got discrimination with some bad

Conrad Meyer (29:16):
Negative

George Mueller (29:17):
Outcomes, that’s got to be protected against. You can’t have that based on race, gender, other protected classes. You do not want to have negative outlaw.

Conrad Meyer (29:26):
Now didn’t Illinois do something too?

George Mueller (29:28):
They did. I think what they did though, I think was in a mental health space. So what they have is they prohibit AI systems from making independent therapeutic decisions or directly interacting with clients in therapy or psychotherapy without having, again, seem very common sense-

Conrad Meyer (29:45):
The human review. … not having an

George Mueller (29:47):
Actual licensed professional do it, right? So what you don’t want to have is effectively they’re prohibiting the unlawful practice of medicine by an AI chatbot or an AI interface, which again, common sense This makes sense, but

Conrad Meyer (30:01):
It’s good to have. So let’s step back for a minute and look at all of what we look at. Yeah, go ahead. On Illinois? Yeah.

George Mueller (30:08):
Yeah, no. I think it kind of buttons up a number of these. What you see here in our context of in negotiating other types of contracts for certain other providers, you see technology creep is what they’re legislating for what it’s now, but also what’s 10 or 20 or 30 years from now. So if you look at, let’s say nursing, like robot nurses or the manner in which technology can say, I don’t say replace nursing, but certainly make easier some of the necessary tasks nursing, just like video monitoring, dispensing medicine, checking blood pressures, checking health physical information. So when you see all this, it says, okay, obviously we don’t want to work in a law that says AI systems can’t have independent therapeutic decisions or directly interact. We think about that now, but think of the movies we’ve seen over

Conrad Meyer (31:08):
The years. I was just thinking of that.

George Mueller (31:09):
So think of the Star Trek, Holodeck and the virtual doctor guy and all that. We talk about it now and we kind of scoff at it, but think about where AI is now and understand what happens with medical technology. We’re not far. And so other than the sophistication of the projected hologram, how far away would we be from actually having some kind of a

Conrad Meyer (31:34):
Virtual

George Mueller (31:35):
Therapeutic interaction?

Conrad Meyer (31:36):
I don’t think we’re that far away.

George Mueller (31:38):
And so I think that’s what this is. Hey, look, we still need human physicians. And that’s what this is stopping right now. So it’s interesting just because I know we look at it immediately and say, “Hey, look, there’s a bull by the horns. They’re trying to fix this. ” I think hopefully they’re making thoughtful legislative decisions about the future, but it’s hard to do that because medical technology is going to hasten and continue to wrap it up.

Conrad Meyer (32:00):
I agree. And so looking at everything that’s going on from Texas to California to Colorado to Illinois at the federal level, okay, ATHS just released or issued request for information asking the public how the Fed should regulate AI and healthcare. I would love to read the comments on that.

George Mueller (32:22):
I’m curious as to who’s going to step in and for what-

Conrad Meyer (32:24):
Oh, they’re going to be comments. Who’s

George Mueller (32:25):
For it? Who’s against it? What do we need to protect against, et cetera, et cetera. So on that point, what should providers do right now? We’ve got four or five states,

Conrad Meyer (32:36):
A

George Mueller (32:37):
Couple of different laws. They’re touching on different components. You and I kind of see them say, “Man, this is practical. It’s common sense. It makes sense. Certainly have some good outcomes, protected characteristics for Colorado purposes. So what should they

Conrad Meyer (32:49):
Do? ” And that’s good. So let’s say this, for our Louisiana providers, if you’re not practicing in Texas, you’re solely in Louisiana. We have not enacted our own AI specific healthcare statutes or regulations just yet.That has not happened, but providers shouldn’t wait. So realize that at a federal level, the enforcement risk from the False Claim Act alone by DOJ that we just discussed, this creates, like you said, a de facto compliance obligation. And if you operate in multiple states and if your patients, especially from telemedicine standpoint, or if you’re submitting claims using an EHR that’s got the nudge, nudge, wink, wink, you better make sure that you’ve got things in check from a regulatory standpoint.

George Mueller (33:30):
Hopefully it’s the nudges without the winks,

Conrad Meyer (33:32):
Right? That’s right.

George Mueller (33:35):
If it wink statute, it’s got real

Conrad Meyer (33:36):
Problems. We’ve got problems.

George Mueller (33:38):
I immediately thought of the emoji winking at you like that. Link. We don’t want that. Again, that’s our brief comic relief here. So I think that to that point from a kind of corporate governance standpoint, even if the state doesn’t have a law, keep in mind that False Claims Act, they’re going to think of any a number of ways to determine what you’ve done after looking at what you’ve done to be false. Even if they don’t have an AI specific law federally, they’re going to say, “Well, you still had a lot of threes that went to fours and a lot of fours that went to fives. And when we evaluate them over the population of claims data, it’s not matching up the diagnostic criteria and the patient outcomes. So you’ve got-

Conrad Meyer (34:17):
You shouldn’t have that much of a jump.

George Mueller (34:18):
Right. And we want to help our listeners and our providers avoid having those things up. So from a corporate governance standpoint, obviously you’re talking about leadership, C-suite executives, healthcare organizations. I think they need to all be looking at this. AI governance is kind of the new cybersecurity. I mean, I think technology and expectations and reasonableness on protection of privacy and avoiding data leaks, which is its own … We could do a whole show on those things. You’ve got to treat AI governance and AI compliance as kind of the same way you look at it from

Conrad Meyer (34:52):
A cybersecurity

George Mueller (34:52):
Standpoint. Kind of top down, obviously visiting an AI use policy and we have issues naturally with privacy and then its use and then being able to control the outcome. I think you and I both know the manner in which you task and command and feed AI directly controls how it works. So you have to have an AI policy. Who can use it, who can’t use it, what forms they can use, who’s allowed to use it, what that data you get from using is allowed to populate where else it can affect your workflow and enter your workflow. And that starts with getting an inventory of every AI tool in your organization, right including the ones from the individuals who are using it kind of on their own. And you probably need some type of human oversight protocols for any AI that’s touching your clinical

Conrad Meyer (35:43):
Decision

George Mueller (35:43):
Making.

Conrad Meyer (35:43):
I agree. Totally agree. And look, this is not … And for you doctors in private practice, this is just a hospital problem. Let me say that. If you’re using an AI tool, whether it’s some sort of diagnostic assistant or an AI scribe, which I know they’re prevalent now in hospitals, they used to have scribes. Now it’s all AI driven. Some coding software, you need to know what that is doing. You cannot just blindly turn the eye and you need to know how that is affecting your documentation and billing on the claims, especially in your EHR and what your disclosure obligations are, especially with FCA. So I think, George, I think what I want to do, and we’re going to start doing this more practical takeaways. We’re going to give our listeners three things they can do right now this week to get ahead of these issues.

(36:29):
Yeah.

George Mueller (36:30):
First thing I said, I’m going to repeat myself. Get an AI inventory, get wherever it touches your workflow. Sit down with your IT team, your technology officer, your compliance officer, your practice administrator, billing manager, and get an inventory, make an actual physical list. You can do it by hand. Of every AI tool, feature, or automated system in your practice, whether it’s embedded, whether it’s a separate service that people are accessing, including AI features that are built in your EHR, obviously, or other things where it’s kind of built in either as a drop-in or an add-in, check your coding software, documentation tools, any kind of high volume computer processes, clinical decision support systems you mentioned earlier, any patient facing chatbots we talked about earlier. You have to have all those kind of touch pointed, kind of a documentation internal control, so to speak, with respect to AI so you can at least begin the inventory and know what to do.

Conrad Meyer (37:31):
That’s a great point. Great point. Now lastly, number two, the second takeaway, implement meaningful human review. We talked about that today, especially for AI generated claims. If you have an AI tool in your workflow, it’s influencing how the claims are coded and submitted. And if you’re using some modern EHR, the answer should be yes, you are having some AI in there as well. You need to document a process for human review so that it’s not just you have someone eyes on, right? Before they go out the door.

George Mueller (37:57):
This is not just set it and forget it.

Conrad Meyer (37:59):
No, it’s not. And the DOJ is made clear that blindly accepting the AI outputs is considered reckless disregard under FCA. So you’re going to be really on the hook for it. So make sure you have a documented review process that is your best defense to FCA claim.

George Mueller (38:12):
Absolutely. And if you’re someone that has telehealth or other things-

Conrad Meyer (38:17):
This is a third takeaway.

George Mueller (38:18):
Yeah, right. Yeah. Number three, know your multi-state obligations.

Conrad Meyer (38:23):
Very important. If

George Mueller (38:24):
You’ve got telehealth or things that be cross jurisdictional or you’re dual licensing in multiple states, specifically right now, Texas, California, Colorado, Illinois, you’ve got new laws on the books, right? So evaluate whether or not those new AI disclosure and governance laws apply to you in any way. And look, if you’re just a Louisiana only practitioner, which a lot of our listeners are, you still need to look at kind of building your AI governance framework as we discussed because you want to be substantively in compliance and I’m sure some of this stuff’s headed our way in one way or the other and it’s a best practice. If someone comes in, sells you an EHR and says, “We’re going to boost your billing by 30%.” Well, the presumption is that, so is your billing really that far off and that wrong right now and it may not be, right?

(39:07):
I mean, some of the promise, efficiency and uptake in billing and collections promised by this could be somewhat distortively the result of some aggressive coding or some upbilling momentum that you want to avoid for exposure purposes.

Conrad Meyer (39:23):
And look, and I’ll add a bonus to our three takeaways, George. If you’re considering any healthcare transaction, buying a practice, bringing on a partner, selling to a private equity group, make sure AI compliance is part of your due diligence checklist, George. You and I see that more and more.

George Mueller (39:40):
And what happens is if you have your sort of like getting ready for a beauty contest metaphorically, but just for the sake of avoiding exposure, if you get ahead of us and have these types of best practices in place, what you’re doing is minimizing the size of the stick that the healthcare regulatory attorney on buyer side is going to hit you with and beat you up with in the diligence when they come out at the last minute and say, “Oh, we’ve got all this exposure and the rep and warrant policy is going to go up by 20% because this exposure is significant or it’s a unique indemnity and we got to reserve additional money for it. “

Conrad Meyer (40:15):
On the hold back.

George Mueller (40:16):
Or

Conrad Meyer (40:16):
Whatever,

George Mueller (40:17):
Right? Or they, God forbid, they say, “Oh, and we’ve got to make a forced disclosure of this. ” And all these things that usually pop in at the end of deals, which are kind of buyer leverage points, right? We want to avoid those.

Conrad Meyer (40:30):
Well, George, I want to thank you. This has been really good. I really enjoyed this episode and I want to tell our listeners if y’all have any questions about AI compliance in your healthcare practice, you need help evaluating your False Claim Act exposure or you just want to develop an AI governance policy, give us a call here at Shohardy Sherman 504-833-5600 or visit us at our website, www.shahardy.com. Our team here at Shahardi Sherman Williams of healthcare providers or works with healthcare providers across Louisiana, the Gulf South, actually across the nation on exactly these issues. And look, we also mentioned this too. If you have a healthcare law topic you’d like us to cover on a future episode, make sure you head on over there to our website. Let us know. We’d love to hear from you.

George Mueller (41:12):
Absolutely. And if you’re thinking of undertaking or in the middle of a healthcare transaction as a seller or buyer or just as a funding source and you want to make sure your AI compliance is covered in your due diligence, we can help with that too. If you find this episode helpful, please subscribe to Health Law Talk whenever you listen Apple Podcasts, Spotify, Google Podcasts, and everywhere else podcasts are available. Leave us a five star rating if you can and review while you’re there. It really does help other healthcare professionals and attorneys find us and we appreciate every single one of them and the feedback contained therein.

Conrad Meyer (41:46):
Absolutely. Look, share this with a colleague. I think George AI in healthcare is not just a future, it’s right now. It’s going on right now. There’s no Star Trek here. We’re doing it every day and every provider needs to be paying attention.

George Mueller (41:58):
Absolutely. Keep an eye out for our next episode. All the changes happening in healthcare regulation right now. We’re not going to run out of topics anytime soon.

Conrad Meyer (42:07):
That’s great. Thank you very much, George, for Health Law Talk. I’m Conrad Meyer.

George Mueller (42:10):
And I’m George Mueller. Thanks for listening.

Introduction (42:14):
Thanks for listening to this episode of Health Law Talk presented by Chehardy Sherman Williams. Please be sure to subscribe to our channel. Make sure to give us that five star rating and share with your friends. Chehardy Sherman Williams is providing this podcast as a public service. This podcast is for educational purposes only. This podcast does not constitute legal advice, nor does this podcast establish an attorney-client relationship. Reference to any specific product or entity does not count as an endorsement or recommendation by Chehardy Sherman Williams. The views expressed by guests on the show are their own and their appearance does not imply an endorsement of them or their entity that they represent. Remember, please consult an attorney for your specific legal issues.