AI 의료 기록 오류로 환자 좌절

Error by AI scribe during medical appointment leaves patient devastated By the Specialist Reporting Team's Paige Cockburn Topic: AI Posted Fri 14 Aug 2026 at 4:55am Fri 14 Aug 2026 at 4:55am Fri 14 Aug 2026 at 4:55am , updated Fri 14 Aug 2026 at 6:04am Fri 14 Aug 2026 at 6:04am Fri 14 Aug 2026 at 6:04am When Rebecca Green arrived for her first appointment with a urologist, she was asked whether she consented to having the appointment transcribed by artificial intelligence (AI). She said yes without hesitation. "I don't want to be in her bad books or be that annoying patient who takes up more of her time," she said. But that decision would cause her a lot of stress, with the AI scribe making up the serious but false claim that Ms Green was taking illegal drugs. "I was gobsmacked and in tears … it made up that I take [psychedelic] mushrooms. I have never done mushrooms in my life," she said. New research by a privacy watchdog has raised serious concerns about AI being used in medical settings due to its propensity to "hallucinate" false symptoms or conditions, and warned patients were at risk. Ms Green only picked up the mistake after her kidney stone surgery in March, when she read the post-operative letter her specialist sent to her GP. It stated Ms Green micro-dosed mushrooms and that could be the reason for prior bleeding around the kidneys. "For this letter to say something that wasn't based in any evidence but could potentially have huge consequences for me … it was quite scary," she said. "It felt very impersonal to me. I wasn't getting the level of care and familiarity from the doctor that I would be hoping for in such a vulnerable situation." Ms Green was particularly stressed as she was receiving workers' compensation and felt any mention of illegal drugs on her medical record could impact her case. Australia's peak regulator of medical practitioners, AHPRA, said clinicians must always check all output from an AI scribe for accuracy to ensure they meet their professional obligations. Ms Green said it was clear her doctor did not do this. "If you had checked that letter you would have been able to correct that before it was sent out … why aren't you in tune with each of your patients and what's going on?" After Ms Green made a complaint, the urologist sent a letter apologising. "Please be assured that we take the accuracy of our documentation very seriously, and we are truly sorry that this occurred," the letter stated. The doctor said she had not been able to determine how the claim about mushrooms came to be included, but it appeared to be from an error during the "dictation or transcription process". The urologist corrected the correspondence and in her letter said she would review how AI was used in her practice. The doctor did not provide any responses to the ABC's questions about what changes, if any, have been made since. AI confuses which breast has cancer Ms Green feels lucky she spotted the error, but is worried some people might not be aware of mistakes on their medical record, which could have repercussions for their health, their career or their insurance. This week the ABC revealed instances where private health insurers were asking providers for their patients' full clinical history as part of "aggressive" auditing practices. AI scribes have become prevalent within the healthcare sector. The Royal Australian College of GPs (RACGP) last year estimated that 40 per cent of GPs were regularly using them, but that number was considered conservative. Research shows having AI transcribe an appointment can reduce the administrative burden for GPs and allow them to focus more on their patient, rather than the computer. But for every benefit, there are risks. A recent report by advocacy group Digital Rights Watch (DRW) found Australians were being exposed to unregulated AI medical scribes, sometimes without their consent, and errors had the potential to change clinical decision-making. Examples of mistakes found in the DRW report included a case where a scribe recorded the wrong breast in a diagnosis of breast cancer and said someone had epilepsy, when they did not. Chief executive of the Consumers Health Forum of Australia Elizabeth Deveny said she often heard complaints of AI mixing up words and patients only finding out by chance. "People report both frustration in finding the original source of the error and then getting the record changed," she said. Dr Deveny, who is currently involved in a major research project on AI medical scribes, said doctors may be too trusting of AI. "There's evidence that for the first few weeks of using AI, [people] check things pretty carefully and then at a point they stop checking because they assume it will be right." Perth GP Sean Stevens, who has been using a scribe for three years, said it was incumbent on the doctor to check every transcript, but believed the technology was becoming more accurate. "Three years ago, I would have to correct multiple things on each transcription. These days it's like one correction every third or fourth transcription." Dr Stevens is a big advocate of tech; he is the RACGP's Digital Health and Innovation chair as well as an advisor for one of the biggest AI scribes, but he admits AI has its weaknesses and can produce incorrect results that sound plausible. "The classic is, it will get the side of the body incorrect and say right when you've said left … so you need to check things pretty closely, especially things like drug doses," he said. He said using AI did not mean he could squeeze more patients in, but it allowed him to run on time more often and produce more thorough referrals. Consenting to AI's involvement The DRW found huge discrepancies in how consent is being obtained, with some doctors only putting a sign in the waiting room, which does not meet privacy law requirements. Other times consent must be given when the appointment is booked online, and alarmingly some patients are even being refused appointments if they do not consent. Dr Deveny said consent must be enthusiastic and fully informed, and patients should feel comfortable asking questions, even if it feels intimidating. The DRW report also identified that people might not be forthcoming about sensitive issues, like family violence, when they know AI is listening. Dr Stevens said that was why he asked for consent in every appointment, even with regular patients. "It's very important patients disclose any and all things that are relevant to their health … and that they know they can say no [to AI] … as a doctor you have to respect that," he said. TGA flags breaches will be dealt with There are around a dozen AI medical scribes available in Australia, but none are approved by the regulator — the Therapeutic Goods Administration (TGA) — because if the product claims to be solely for transcription and summarising, it is exempt from oversight. The TGA only considers tools that do things like suggesting a diagnosis or treatment plan as a "medical device". Currently, no AI scribes have gone through the regulation process, despite some claiming to offer clinical suggestions. The TGA has acknowledged this means some scribes were potentially in breach of the law . The TGA has spent the last year reviewing AI scribes and says its findings will be released in the coming weeks, and any breaches will be acted upon. Chief executive of Patients Australia Lisa Robins believed all scribes should be regulated, as summarising a patient still involved decision-making by AI. Dr Stevens had concerns that extending regulation would only stifle future innovation. The DRW is calling on the newly established AI Safety Institute to test AI scribes and publish a list of approved tools. This would be similar to New Zealand's approach, where the national AI advisory group reviewed tools and last year gave approval for clinicians to use two AI too