AI scribes increasingly adopted by New Zealand GPs despite ethical, legal, and data security concerns

Many general practitioners in New Zealand are adopting AI scribes to transcribe patient notes during consultations, despite unresolved ethical, legal, and data security concerns, according to a study by the University of Otago, Wellington—Ōtākou Whakaihu Waka, Pōneke. The research, published in the Journal of Primary Health Care, surveyed 197 primary care providers—including GPs, nurses, nurse practitioners, rural emergency care providers, and practice managers—between February and March 2024. The findings reveal that 40% of respondents were already using AI scribes, with many reporting significant time savings—47% estimated that using the tools in every consultation could free up between 30 minutes and two hours per day. However, a notable portion found that editing and correcting AI-generated notes consumed so much time that no net time gain was achieved. While many practitioners found AI scribes helpful for improving patient interaction—allowing more eye contact and active listening—others expressed serious concerns. Accuracy and completeness of notes were major issues, with some doctors reporting that critical negative findings were missed. One GP noted that the software’s “hallucination rate” was high and often subtle, leading to mistrust. Others highlighted difficulties with understanding New Zealand accents and te reo Māori, prompting some to pause recordings when discussing sensitive patient information like names or dates of birth. The use of AI scribes also altered the consultation dynamic, requiring doctors to verbalize their physical exam findings and clinical reasoning aloud to ensure the tool captured everything. One GP described having to say out loud, “Oh, pain in the right upper quadrant?” after a patient pointed to an area of discomfort. Ethical and legal concerns were widespread. Only 66% of users had read the software’s terms and conditions, and just 59% reported seeking patient consent. Professor Angela Ballantyne, a bioethicist and lead researcher, emphasized that clinicians remain legally and professionally responsible for the accuracy of patient records, regardless of whether AI was used. Yet, the time required to verify AI-generated notes often offset any efficiency gains. Data security remains a pressing issue. Most AI scribes rely on international cloud platforms, raising concerns about data storage, access, and protection from cyber threats. There are also unique Aotearoa-specific challenges, particularly around Māori data sovereignty and governance. In July, Health New Zealand—Te Whatu Ora’s National Artificial Intelligence and Algorithm Expert Advisory Group endorsed two ambient AI scribe tools, Heidi Health and iMedX, for use in the country after assessing their privacy, security, and ethical implications. Professor Ballantyne stressed that the use of AI in healthcare should not be assumed to have patient consent. Patients must be informed and given the right to opt out without affecting access to care. She also called for clear training, robust guidelines, and updated regulatory frameworks. The Medical Council of New Zealand is expected to release guidance on AI use in healthcare later in the year, likely requiring explicit patient consent. While AI tools are improving, Professor Ballantyne noted that with proper training, governance, and patient engagement, AI scribes could still play a valuable role in the future of primary care.