Adoption of AI scribe in Australian healthcare: concerns about errors in records and protection of patient data

According to the Off the Record report by Digital Rights Watch, more than 40 % of Australian doctors use AI scribe tools, which record consultations and use large language models (similar to ChatGPT or Claude) to generate notes for medical records. This is an attractive solution for doctors because it saves hours of administrative work, and the Australian Medicare system only reimburses time spent directly with the patient anyway, not writing notes.

However, the report highlights risks to patients. These tools often do more than transcribe speech: they also interpret and summarise the conversation based on what the AI concludes the doctor would say — while hallucinating, meaning they confidently record things that did not happen. According to the report, AI systems also exhibit biases linked to race, ethnicity, class, and gender, which can lead to records that differ significantly from reality for patients from culturally and linguistically diverse backgrounds. Older studies reported that up to 90 % of AI notes required correction, while newer studies report approximately 20 % of notes containing errors serious enough to affect the diagnosis.

The report also points to a lack of transparency: according to its findings, healthcare professionals cannot tell patients specifically where their data goes, who has access to it, or which AI models are used — they receive only general assurances that the data stays in Australia. Digital Rights Watch therefore proposes that the Australian regulator for medicines and medical devices (TGA) tighten regulation of AI scribes and treat them like any other medical device subject to safety testing — TGA currently considers a scribe to be a medical device only if it provides diagnostic recommendations, which, according to the report, allows providers to avoid scrutiny.