According to a new study from the American Medical Association (AMA), 81 percent of physicians reported using artificial intelligence (AI) tools or systems at work in 2025, more than double from when the AMA first polled physicians on health AI in 2023. Among the most quickly adopted AI applications has been medical scribes  —  AI systems designed to automatically document patient encounters, generate clinical notes, and assist with medical documentation. ¹

“These AI assistants can reduce a physician’s time devoted to documentation by up to 70 percent by transcribing patient encounters, entering data into EHRs, and processing information for orders and prescriptions, allowing physicians to focus on direct patient care.” ²

Now that AI scribes have become commonplace in clinical settings, the same AMA study shows that physicians are concerned with the risks AI presents to data privacy and the integrity of the physician-patient relationship, specifically if patients use AI to interpret test results or other tasks that require clinical judgment. ¹

The benefits of AI scribes

‍AI scribes have become popular due to increased documentation demands over the past decade. In 2024, the AMA found that physicians spend an average of “5.8 hours per eight hours of time scheduled for patients on documentation duties, with a significant amount of that time spent outside of scheduled clinical hours.” These demands contribute significantly to professional frustration and burnout. ³

AI scribes use “machine learning” (a type of AI that allows systems to learn from data without human programming or prompts); “natural language processing (NLP)” (AI that enables systems to interpret and use human language); ⁴ and other capabilities (such as Large Language Models, trained on huge amounts of formal and informal text) to:

• record and summarize patient-physician conversations;
• extract relevant medical information from discussions;
• input data directly into EHR systems; and
• recommend medical codes for billing and compliance purposes.

Physicians can then review, edit, and sign the AI-generated notes in the medical record. This last step is vital to avoid documentation errors, omissions, or misrepresentations of the encounter. Skipping this step could increase medical liability risk.
‍

Caveats for physicians using AI medical scribes

1. Documentation inaccuracies and omissions

AI medical scribes are certainly not perfect. These systems, also called “ambient medical scribes”  translate recorded interactions and input them into the patient record. These systems may:

• miss critical clinical information discussed with the patient;
• incorrectly transcribe medical terminology;
• omit important non-verbal observations;
• “hallucinate” or create information about a patient simply to provide an answer in an EHR template or fill in an information gap; ⁵‍
• misinterpret vague statements or questions; or
• omit contextual factors that influenced clinical decision-making.

There are also concerns about AI scribes relying on generalized, historical data the system may encounter that perpetuate racial, sexual, age-based, and other biases. When past biases are introduced into the AI system, there is a potential for discriminatory practices or less inclusive treatment. ^{6, 7}

2. Reduced critical review

As physicians and other health care professionals grow more accustomed to using AI-generated documentation, there may be a temptation to spend less time reviewing what has been recorded. This may result in signing off on notes without careful review; not catching mistakes in the patient record; making assumptions that important information was captured; or not applying appropriate critical thought to what was documented.

AI systems, even those with high accuracy rates, can make mistakes, especially in complex cases. This makes physician review imperative. “Uncritical acceptance of AI suggestions carries the risk of errors in the chart.” ²

3. Concerns regarding legal authorship and authentication

When using an AI scribe to generate a physician’s notes, consider the following legal questions.

• Who is legally responsible for the AI-generated content?
• How are physician observations and AI-generated content differentiated?
• How appropriate or “safe” are automatic signatures on AI-generated content?

Guidance from the Federation of State Medical Boards states that physicians remain fully responsible for the content of all medical documentation, regardless of how it was generated. Therefore, physicians must take the time and care to ensure AI-generated content and patient records accurately reflect their professional assessment of the patient, care goals, and the treatment provided. Automatic signatures on AI-generated content are discouraged. ⁸

The BMJ adds, “As a rule of thumb, the higher the stakes of the AI output, the greater the level of scrutiny from the supervising clinician.” ⁹

4. Privacy breaches and HIPAA concerns

Because AI medical scribe notes are based on audio recordings of physician-patient conversations, there is a potential for breaches of patient privacy. These conversations may include mentions of HIPAA-protected patient information, such as patient name, birthdate, address, contact information, family member names and contact information, pre-existing conditions, or other protected health information (PHI). In turn, this information becomes vulnerable in the event of a cyber security breach; inappropriate storage or transmission of PHI; or sharing PHI with third-party AI vendors.

The Office for Civil Rights (OCR) maintains that health care providers covered by HIPAA are responsible for mitigating or preventing cyber-threats to patient PHI. ^{10, 11} Before choosing an AI scribing software, review its protocols and security measures to ensure HIPAA compliance. In addition, a Business Associates Agreement (BAA) with the vendor should be executed. If AI scribing is built into your current EMR system, verify that the integrated AI scribing portion is not offered by a separate vendor. If it is managed by a separate vendor, consider obtaining a separate BAA. ^{11, 12}

5. Informed consent challenges

Patient consent for the use of AI in their care and treatment has raised several questions being addressed by lawmakers across the U.S.

• Are patients adequately informed that AI is recording and processing their conversation, symptoms, and diagnoses?
• Do patients understand how their PHI will be used and analyzed by the AI system?
• Have patients been given the opportunity to opt out of having an AI scribe or system used in their care?
• If applicable, do your consent forms provide information on how patient data may be used by AI to “train” the system?

Texas law now requires physicians to disclose that they are using AI in patient treatment.

The Texas Responsible AI Governance Act requires providers to inform patients or guardians when AI is used in treatment, no later than the date of service or as soon as reasonably possible in emergencies. Disclosures must be clear, conspicuous, and in plain language. This law must be followed alongside existing HIPAA requirements.

Clear disclosures of AI use should include:

• what AI tools are used;
• how the tools are used, noting that providers must review the AI outputs before acting on them;
• the data — including PHI — being processed and how it will be protected; and
• a way patients can ask questions or raise concerns about the AI use.

Any disclosure language should be adapted to fit your practice and specific use of AI. Samples of clear disclosure language can be found here.

Physicians may wish to consult with an attorney if their use of AI is complex, or if further assistance is needed to develop disclosure language. ¹³

Several other states require explicit consent from patients for the recording and processing of health care encounters using AI. Failure to obtain consent could lead to claims alleging violations of patient privacy. ¹⁴

6. Risks of using clinical support features

Some advanced AI scribe systems include clinical support features. These features carry additional risk for a physician, including:

• relying on AI clinical suggestions without careful review and agreement;
• unclear documentation about which recommendations came from AI and which came from the physician; and
• errors in AI clinical suggestions based on the system’s limited information or training.

A recent study comparing different scribing tools found that omission errors were common. "Across platforms and studies, omission errors account for roughly 71 to 83 percent of all errors. Many omissions are clinically minor, but a concerning number are not." While an AI scribe may produce a patient note that is thorough and well-organized, “it may quietly drop a critical detail about disease severity, a comorbid condition, or a nuance in the differential diagnosis or management plan.” ¹⁵

These types of risks may be hard to defend in the event of a claim due to a misdiagnosis or other diagnostic error traced to AI use.
‍

Risk management strategies when using AI scribes

‍Consider the following steps when using AI medical scribes. ^{7, 16, 17, 18, 19, 20}

1. Establish strict review protocols

• Revise your current documentation policies and procedures to include review of all AI-generated documentation and content.
• Set aside dedicated time each day to carefully review documentation and make any needed edits prior to sign-off.  
• Whenever possible, review the notes promptly while the patient encounter is still fresh.
• Reconcile the notes with the actual actions of the visit: medication, lab work, or imaging orders, referrals, procedure documentation, patient instructions given, and follow-up plan.
• After completing the notes, read it back as if you were potentially reading it as the patient’s next clinician and as the patient themselves. Update any ambiguities and make any corrections.
• If an error is noted after the AI-generated note has been finalized and signed, correct it via an addendum or amendment, which clearly shows the corrections made, including name of physician or provider reviewing the notes, date of review and error found. Follow applicable state rules regarding documentation. For example, the Texas Medical Board rule 163.1 requires “clear identification of any amendment or correction to the medical record, including the date it was amended or corrected and the identity of the author of the amendment or correction, with the original text remaining legible.” ²¹

2. Include AI scribe use in consent forms and procedures

• Add language to your patient consent forms that addresses AI scribe use. Explain the technology and its purpose in your informed consent discussions with patients using clear, non-technical language.
• Train staff on how to clearly explain AI documentation processes to patients.
• Document all consent discussions in the patient’s record.

3. Train staff on AI scribe/documentation use

• Ensure all physicians and staff receive thorough training on the use and risks of AI scribes.
• Consider training needs based on role and responsibility of each team member; assess comfort levels among staff; and identify documentation “pain points” to address fully in training and in policies and procedures.
• Regularly update training as AI systems are upgraded or modified.
• Document physician and staff completion of AI scribe training, including training dates and levels for each staff member.

4. Adopt “built-in” technical safeguards in your EHR system

When using AI scribes that interact with your EHR, explore and use any built-in safeguards for your EHR, such as:

• prompts for a physician to confirm or verify important information, such as medications, dosages, or patient allergies;
• highlighting or “flagging” potential inaccuracies for physician review; or
• requiring physician review of AI-generated content before closing a record.

5. Stay informed about the use of AI in health care

• Keep track of evolving legal and regulatory guidance for AI use in health care.
• Understand how clinical information captured by the AI scribe may be stored, retained, accessed, and subsequently used.
• Consider taking professional development courses or training in appropriate and legal AI use. Document any training or education you receive on AI use.
• Contact your medical professional liability insurer about any AI-specific concerns you may have.
  ‍

Conclusion

AI medical scribes offer significant benefits in reducing administrative burden and potentially reducing physician stress and burnout. However, their use introduces new liability concerns. Physicians using these tools must remain vigilant in their oversight of these systems and their documentation; establish strong policies and procedures for their safe use; ensure appropriate patient informed consent procedures; and stay informed about developing legal standards for AI use.
‍

Sources

1. O’Reilly KB. More than 80% of physicians use AI professionally: AMA survey. American Medical Association. March 12, 2026. Available at https://www.ama-assn.org/practice-management/digital-health/more-80-physicians-use-ai-professionally-ama-survey. Accessed September 17, 2026.
2. Reddy Bonguraia A, Save D, Virmani A, et al. Transforming Health Care with Artificial Intelligence: Redefining Medical Documentation. Mayo Clinic Proceedings: Digital Health. Volume 2, Issue 3. September 2024. Available at https://www.sciencedirect.com/science/article/pii/S2949761224000415.  Accessed September 17, 2026.
3. Robeznieks A. Five physician specialties that spend the most time in the EHR. American Medical Association. September 11, 2024. Available at https://www.ama-assn.org/practice-management/digital-health/five-physician-specialties-spend-most-time-ehr. Accessed September 17, 2026.
4. Barth S. AI in Healthcare. Foresee Medical. 2018. Available at https://www.foreseemed.com/artificial-intelligence-in-healthcare. Accessed September 17, 2026.
5. Canadian Medical Protective Association. AI Scribes: Answers to frequently asked questions. Revised December 2025. Available at https://www.cmpa-acpm.ca/en/advice-publications/browse-articles/2023/ai-scribes-answers-to-frequently-asked-questions. Accessed September 17, 2026.
6. Hasanzadeh F, Josephson CB, Waters G, et. al. Bias recognition and mitigation strategies in artificial intelligence healthcare applications. NPJ Digital Medicine. Nature. March 11, 2025. Available at https://www.nature.com/articles/s41746-025-01503-7. Accessed September 17, 2026.
7. American Medical Association. Augmented Intelligence Development, Deployment, and Use in Health Care. November 2024. Available at https://www.ama-assn.org/system/files/ama-ai-principles.pdf. Accessed September 17, 2026.
8. Federation of State Medical Boards. Navigating the Responsible and Ethical Incorporation of Artificial Intelligence into Clinical Practice. Adopted April 2024. Available at https://www.fsmb.org/siteassets/advocacy/policies/incorporation-of-ai-into-practice.pdf. Accessed September 17, 2026.
9. Sokol D. Six rules for doctors using AI. The BMJ. September 17, 2026. Available at https://www.bmj.com/content/394/bmj-2026-100883. Accessed September 17, 2026.
10. U.S. Department of Health and Human Services. HHS Office for Civil Rights Settles HIPAA Security Rule Investigation with Northeast Radiology. Press release. April 10, 2025. Available at https://www.hhs.gov/press-room/hhs-ocr-hipaa-settlement-nerad.html. Accessed September 17, 2026.
11. PrivaPlan. Ambient AI scribe and HIPAA compliance: What every healthcare clinic needs to know (2026). April 9, 2026. Available at https://privaplan.com/ai-ambient-scribes-is-your-health-care-clinic-ready/. Accessed September 17, 2026.
12. Ummer S. Ambient AI scribes – Efficiency gains vs. emerging privacy and cybersecurity risks. American Bar Association. February 10, 2026. Available at https://www.americanbar.org/groups/health_law/news/2026/ambient-ai-scribes-privacy-cybersecurity/. Accessed September 17, 2026.
13. Brockway LH. Texas physicians must disclose use of AI under new state rules. The Reporter, Quarter 1, 2026. Texas Medical Liability Trust. Available at https://www.tmlt.org/resource/texas-physicians-must-disclose-use-of-ai-under-new-state-rules. Accessed September 18, 2026.
14. Good C. Sample Patient Consent Form for Using Artificial Intelligence for Dictation, Transcription. Medical Group Management Association. April 2, 2024. Available at https://www.mgma.com/member-tools/sample-patient-consent-form-for-using-ai. Accessed September 17, 2026.
15. Pines JM. AI scribes for doctors are everywhere. Here’s what the research says. Forbes. Updated March 25, 2026. Available at https://www.forbes.com/sites/jessepines/2026/03/06/medical-ai-scribes-are-everywhere-research-shows-benefits--risks/. Accessed September 17, 2026.
16. Henry TA. 7 tips for responsible use for health care AI. American Medical Association. March 4, 2021. Available at https://www.ama-assn.org/practice-management/digital-health/7-tips-responsible-use-health-care-ai. Accessed September 17, 2026.
17. Vanderpool D. Risk Management: Artificial Intelligence in Clinical Practice. Innovations in Clinical Neuroscience. December 1, 2024. Available at https://pmc.ncbi.nlm.nih.gov/articles/PMC11709444/. Accessed September 17, 2026.
18. Agarwal P, Lall R, Girdhari R. Artificial intelligence scribes in primary care. Canadian Medical Association Journal. September 16, 2024. Available at https://pmc.ncbi.nlm.nih.gov/articles/PMC11412733/. Accessed September 17, 2026.
19. OrbDoc. Review AI-Generated Clinical Notes Before Signing. July 18, 2026. Available at https://orbdoc.com/blog/review-ai-generated-clinical-notes-before-signing. Accessed September 17, 2026.
20. Purohit G. Best Practices for Training Clinical Staff on AI Medical Documentation. Omni Digital Healthcare Innovators. January 16, 2026. Available at https://omnimd.com/blog/ai-medical-documentation-training-best-practices/. Accessed September 17, 2026.
21. Texas Administrative Code. Rule 163.1 (b) (7). Available at https://texas-sos.appianportalsgov.com/rules-and-meetings?$locale=en_US&interface=VIEW_TAC_SUMMARY&queryAsDate=05%2F29%2F2025&recordId=223450. Accessed September 17, 2026.

‍

Updated September 18, 2026, from previous version published in the Reporter, Quarter 2 edition, 2025.