The artificial intelligence market in healthcare is projected by some analysts to approach $50 billion, and that level of investment is accelerating adoption across clinical workflows. Nowhere is that acceleration more visible than in medical record documentation, where ambient scribe technology—AI tools that listen to patient-physician conversations and generate suggested notes and billing codes—has quickly become one of the most widely deployed AI applications in medicine.

According to a Medical Group Management Association survey, 42% of medical groups are already using ambient scribing technology. Physicians typically activate these tools through their phones, room microphones, or microphones embedded in their laptops. But with rapid adoption comes a complex mix of regulatory, legal, and patient safety considerations that healthcare organizations must address proactively.

“When we talk about AI in the healthcare marketplace from a medical malpractice perspective, we’re looking at how it’s embedded in clinical workflows,” said Laurette Côté, MBA, CPHRM, Senior Risk Management Consultant at ProAssurance, Part of The Doctors Company. “There’s not one unified framework at this point, so it can be difficult for organizations to juggle the overlapping legal requirements.”

The Case for Ambient Scribes—and the Growing Risks

Laurette Côté, MBA, CPHRM, Senior Risk Management Consultant, ProAssurance

The advantages of ambient scribe technology are clear, which explains its rapid uptake. The primary benefit is a reduction in the time physicians spend documenting patient encounters, which allows them to spend more time with patients and less time charting after hours.

That reduction has downstream benefits for one of healthcare’s most persistent challenges: physician burnout.

“The reduction in documentation time can lead to lower levels of frustration and burnout,” Côté said. “Ultimately, the use of ambient scribing, if done appropriately, helps physicians, helps patients, and also helps organizations because lower physician burn out contributes to a decrease in liability risks.”

The American Medical Association has framed AI in healthcare as “augmented intelligence,” emphasizing that AI is a support tool and should never replace physician or clinician decision-making. The AMA recently developed a new policy that addresses the development and implementation of AI in healthcare. While that stance is not binding policy, it can shape future legal narratives and provides a guidepost for risk management expectations.

Even so, the risks are real. Among the most significant is automation bias. That’s the tendency of clinicians to rely too heavily on AI-generated output as it becomes integrated into daily workflows.

“The more you use an automated system, the more the tendency to over-rely on it exists,” Côté said. “We always go back to the example of GPS navigation. All of us use it, but sometimes it brings you to the wrong place.”

When that dependency creeps into healthcare, a diminished review of AI-generated notes can lead to inaccuracies, omissions, or mischaracterizations of a clinical encounter. Those errors can fuel allegations of failure to diagnose in a timely manner or delays in treatment. These are among the most common medical malpractice allegations across physician specialties.

Discoverability is another emerging concern. Raw audio recordings that feed into medical record documentation may be stored in a vendor’s software system, and if a medical record is discoverable during litigation, recordings may be discoverable.

“The intent is to have the physician review the recording for accuracy and then put it in the medical record. However, that recording and that medical record encounter note may slightly differ,” Côté said. “In the world of litigation, even those seemingly minor inconsistencies really do have an impact.”

Patient consent is another area drawing legal attention. Most states follow a one-party consent standard for recordings, but Illinois, California, Florida, and Pennsylvania require two-party consent. A pending case in California involving Sharp HealthCare alleges that patient-physician conversations were secretly recorded using ambient listening systems, and that the AI system automatically inserted consent language into the medical record even when no consent had been obtained.

“If patients aren’t even aware that an AI system is being utilized, they don’t have the ability to opt out,” Côté said. “Lack of disclosure and lack of opt-out rights are fueling some of the cases at this point.”

Practical Steps for Ethical, Lower-Risk Implementation

While the risks are significant, they are manageable when organizations take a deliberate approach to disclosure, vetting, and oversight. Côté recommends focusing on several key areas.

Disclosure should be straightforward. Rather than treating ambient listening notification as a formal informed consent process linked to medical treatment, organizations can use a simpler disclosure standard. That’s often a brief verbal notification supplemented by posted signage.

“Most of our physicians that we’ve interviewed simply get verbal consent by saying, ‘I’m going to use my phone to create the note, and I’d like your consent. This helps me create the medical record documentation that you can further review at any time through the portal,'” Côté said. “And then really importantly, they must document it.”

Vendor vetting is equally critical. Organizations should understand how a system captures information, how it summarizes clinical encounters, how it integrates with the electronic medical record, and how easily corrections can be made. Questions about background noise, accents, and specialty-specific terminology should be addressed up front.

The vetting team should include clinicians, health information management professionals, and information security experts.

“Without input from end users like physicians and healthcare professionals, you won’t understand the practical application needed to avoid risks such as delayed care, inaccurate charting, inappropriate billing, and unsupported orders,” Côté said.

Organizations also need to know exactly what vendors are doing with patient information. If vendors are using protected health information to train large language models, that may go beyond what HIPAA considers permitted use for treatment and operations—and a more formal consent process may be required.

“These issues need to be handled up front at an organizational level, utilizing legal counsel to ensure all the i’s are dotted and the t’s are crossed,” Côté said.

Finally, physicians must remember that the human still owns the note. Providers are required to review AI-generated documentation, confirm its accuracy, and finalize it before it enters the medical record. That review is both a patient safety safeguard and a critical line of defense against professional liability claims.

Partnering with Insurers to Navigate What’s Next

As AI continues to expand beyond documentation into radiology, cardiology, and other clinical specialties, healthcare organizations need partners who understand the intersection of technology, regulation, and liability. ProAssurance supports physicians and healthcare organizations through risk management consulting, CME seminars, practice management resources, podcasts, and published guidance on emerging issues such as AI and privilege during litigation.

“AI as a whole is much larger than just ambient listening systems,” Côté said. “We always try to remind physicians that this isn’t something that can be pushed off. They need to keep up on it from a regulatory standpoint, from a risk standpoint, and from a patient satisfaction and patient safety standpoint.”

For organizations willing to invest the time in thoughtful implementation, ambient scribes offer real benefits, those being more meaningful patient interactions, less after-hours charting, and reduced burnout. The organizations that pair those benefits with strong disclosure practices, careful vendor vetting, and consistent physician oversight will be best positioned to capture the upside while managing the liability exposure that comes with it.

To learn more, visit: RiskManagement.ProAssurance.com.

This article was produced by the R&I Brand Studio, a unit of the advertising department of Risk & Insurance, in collaboration with ProAssurance. The editorial staff of Risk & Insurance had no role in its preparation.