澳大利亚超过40%的医生正在使用AI医疗助手记录患者咨询内容[1]。然而，这些工具在准确性和数据保护方面存在重大问题，引发了对患者隐私和医疗记录质量的担忧。
AI医疗助手的错误率令人担忧[1]。早期研究表明，最多90%的AI生成医疗记录需要修正，而近期研究显示约20%的AI临床记录包含足以影响患者诊断的错误[1]。这些工具还容易产生幻觉、引入性别和种族偏见[1]。
数据隐私和监管透明度同样缺失[1]。AI医疗助手通常未能披露数据处理方式、访问权限和所用AI模型，导致患者对其医疗数据的访问、使用和保护方式缺乏清晰了解，医疗专业人士也无法向患者提供具体的数据隐私保证[1]。澳大利亚治疗用品管理局（TGA）目前仅在AI提供诊断建议时才将其归类为医疗设备，这一做法允许供应商规避审查[1]。Digital Rights Watch建议TGA应像监管其他医疗设备一样监管AI医疗助手[1]。
More than 40% of Australian doctors are deploying artificial intelligence assistants to document patient consultations, yet significant risks accompany this widespread adoption [1]. While these tools promise to streamline clinical workflows, emerging evidence reveals substantial accuracy and privacy challenges that warrant urgent scrutiny.
Early research indicates that up to 90% of AI-generated medical records require corrections, with more recent studies showing approximately 20% of AI clinical notes contain errors substantial enough to affect patient diagnoses [1]. Beyond accuracy issues, AI medical scribes frequently exhibit hallucinations and introduce gender and racial bias into clinical documentation [1]. Patients remain largely unaware of how their medical data is accessed, used, and protected, as healthcare professionals cannot provide concrete privacy guarantees about these systems [1]. Most critically, AI medical assistant vendors typically fail to disclose their data processing methods, who has access to information, or which AI models power their tools [1].
The regulatory landscape compounds these concerns. Australia's Therapeutic Goods Administration currently classifies AI systems as medical devices only when they provide diagnostic recommendations, allowing vendors to circumvent oversight in other applications [1]. This narrow regulatory framework leaves numerous AI scribes operating without formal evaluation. Digital Rights Watch has recommended that the Therapeutic Goods Administration regulate AI medical assistants with the same rigor applied to other medical devices [1].
