Against this backdrop, digital scribes — AI-powered tools that generate clinical documentation — are emerging as a promising solution to reduce the administrative burden on healthcare providers. Digital scribes, like DeepScribe and MEDITECH, typically function by recording the clinician-patient interaction, transcribing the audio into text, extracting key clinical information, and summarizing the encounter into a structured note for the medical record.11 These tools are often embedded within or integrated into electronic health record (EHR) platforms, offering the potential to improve physician efficiency and “presence” during clinical encounters.
In Canada, such applications are considered “lower risk” from a regulatory perspective. The CMPA categorizes administrative AI, including digital scribes, as posing fewer immediate clinical dangers than diagnostic tools.12 Similarly, the College of Physicians and Surgeons of British Columbia (CPSBC) does not impose specific restrictions on digital scribe use, requiring only that patients provide informed consent if their encounters are to be recorded for the purposes of AI-driven documentation.13
Proponents of digital scribes have endorsed these tools as a way to rebalance the demands of modern practice. In a recent article published by the University of British Columbia, three physicians and a medical student praised the use of such “conversational AI” in family medicine. They argued that such tools do not replace physicians but rather “empower them to focus on their primary mission [of] delivering exemplary care,” by alleviating the “burden of documentation” that contributes to physician burnout and inefficiency.14 Similarly, a Canadian study examining the needs of emergency medicine physicians found that “automated charting” and “report generation” were the top-ranked AI applications of interest.15 On a much larger scale, Fraser Health, a regional public health authority in BC, is currently part of a pilot project that uses Google Cloud’s generative AI tools to produce an initial draft of a patient’s hospital course and discharge summary.16 These examples signal a growing institutional and professional enthusiasm for AI-driven documentation solutions.
Medical Records Produced by Digital Scribes: Potential Medical Malpractice Implications
Sections 3-5 to 3-7 of the Bylaws pursuant to the Health Professions Act impose obligations on physicians regarding medical record management.17 Yet, despite the growing integration of AI in clinical documentation, neither the Bylaws, nor guidance from the CMPA or the CPSBC, currently requires that records generated or assisted by AI be explicitly identified as such. In other words, there is no obligation to annotate such documentation as AI-assisted.
This lack of transparency poses significant concerns, particularly because AI-generated content may describe a clinical encounter that did not occur as documented. This issue is not limited to instances of AI “hallucination” — a phenomenon where AI presents incorrect information as accurate.18 Even when the AI-generated summary appears accurate, it may subtly misrepresent the flow or emphasis of the interaction. In extracting key clinical information, it may overlook signs and symptoms that ultimately end up being important. In such cases, the resulting documentation risks becoming a deceptive fiction rather than a reliable record. For example, in the case of Pinch v Morwood, the pregnant plaintiff attended hospital complaining of neck pain which extended into her head but which she described as “neck pain.”19 One of the issues in the case was whether the physician was negligent for failing to elicit that her neck pain included headache, a red flag for pre-eclampsia which ultimately led to an eclamptic seizure and a brain injury to the infant. In similar circumstances, if an AI-generated scribe was used to summarize the encounter without the medical sensitivity to understand the implications of head pain for this particular population subset, and fail to note extension of pain into the head, it would have serious implications for any subsequent legal proceedings.
