Australia’s federal health department has raised concerns about the soaring use of AI scribes by doctors, as the health regulator considers whether new safeguards are needed to protect patient data during GP consultations.
The Guardian reported on July 5 that AI scribe tools — which record, transcribe, and summarise conversations between doctors and patients — are becoming “fast becoming popular in GP surgeries,” prompting regulators to monitor the technology and its potential pitfalls.
🔍 THE BOTTOM LINE
AI scribes are now routinely recording patient consultations in Australia, and the federal health department is worried enough to flag it publicly. New Zealand GPs are using the same tools, with no equivalent government warning. The trans-Tasman regulatory gap on medical AI is about to become a patient-trust problem.
What AI Scribes Actually Do
An AI scribe sits in the consultation — either as a phone app on the doctor’s desk or integrated into practice management software — and listens to the entire appointment. It transcribes the conversation, extracts the relevant clinical information, and generates a summary the GP can paste into the patient’s electronic health record.
The appeal is obvious. GPs spend an estimated 15-30% of their day on documentation. An AI scribe that handles that load means more face time with patients, less after-hours admin, and shorter wait times. The technology works — the transcription is accurate, the summaries are clinically useful, and doctors who use it report meaningful time savings.
The problem is everything else. The recording captures the entire consultation, including sensitive disclosures that may not be clinically relevant but are deeply personal. The transcription is processed by a third-party AI provider, often offshore. The summary becomes part of the medical record, but the raw recording and transcript may be retained by the scribe vendor for model training or quality assurance. Patients are rarely asked for meaningful consent — a sign on the waiting room wall is the typical standard.
What the Australian Government Is Worried About
The Guardian’s reporting is based on an exclusive investigation that found the federal health department has “raised concerns about the use of AI scribes by doctors as the health regulator considers the need for safeguards around the technology.”
The specific concerns are the ones you’d expect: where is the recording stored, who has access to it, how long is it retained, is it used for model training, and what happens when a patient withdraws consent. The Australian health regulator is “monitoring its implementation and potential pitfalls” — which is regulator-speak for “we know this is happening faster than we can regulate it.”
Australia already has the AI disclosure deadline framework and has been more aggressive than New Zealand on AI governance across multiple sectors. The fact that even Australia’s regulators are playing catch-up on medical AI tells you how fast this market is moving.
Why New Zealand Should Care
New Zealand GPs are using the same AI scribe tools. The products are international — ScribeSense, Heidi, Lyrebird, and similar vendors operate across both markets. The clinical workflows are identical. The privacy risks are identical.
What’s different is the regulatory signal. Australia’s health department has at least flagged the concern publicly. New Zealand has no equivalent warning, no guidance for GPs, and no framework for patient consent in AI-recorded consultations. The Office of the Privacy Commissioner has been active on AI more broadly — the sovereign AI infrastructure debate has surfaced privacy concerns — but medical AI scribes specifically have not been addressed.
The trans-Tasman gap matters because NZ patients assume (reasonably) that if a technology were dangerous enough to warrant a government warning, someone would have said so. Australia’s warning creates an information asymmetry: Australian patients will now be told their GP is using an AI scribe and can ask questions. New Zealand patients won’t know to ask.
The Consent Problem Nobody Has Solved
The hardest question is not technical — it’s consent. A GP appointment involves some of the most sensitive information a person will ever share: mental health struggles, reproductive decisions, substance use, family conflicts, trauma. The current consent model for AI scribes is roughly: a sign in the waiting room, or a line in the practice’s privacy policy that nobody reads.
That model doesn’t work for AI scribes because the stakes are different. A human medical scribe (a trained professional in the room) operates under professional codes and employment contracts. An AI scribe sends the audio to a vendor’s cloud, processes it through a model that may retain the data, and generates a summary with no human oversight of the intermediate steps. The patient has no way to know what happened to their recording between the consultation and the final note in their file.
Australia’s regulator is now grappling with this. New Zealand’s hasn’t started.
What Should Happen Next
For Australia: the health department’s concern needs to become guidance, fast. GPs need clear rules on consent (opt-in, not opt-out), data retention (minimum necessary, not vendor-default), and vendor transparency (where the data goes, what it’s used for). The technology is not the enemy — but unregulated adoption in a clinical setting is a trust time bomb.
For New Zealand: the Ministry of Health and the Privacy Commissioner should issue a joint advisory now, not after the first breach. The Australian investigation gives NZ regulators a head start — use it. The same tools are in the same clinics. The same patients are exposed to the same risks. The same safeguards are needed.
For patients: ask your GP if they’re using an AI scribe. Ask where the recording goes. Ask how long it’s kept. You have the right to say no — and if enough patients say no, the market will respond with better consent practices.
❓ FAQ
Are AI scribes safe? The transcription and summarisation technology is generally accurate and clinically useful. The safety concern is not the AI’s output — it’s the data handling. Where the recording goes, who can access it, and how long it’s retained are the questions that matter.
Can I refuse to have my consultation recorded? Yes. You have the right to refuse AI scribe recording under both Australian and NZ privacy law. If your GP uses an AI scribe, they should offer you the option to opt out before the consultation begins.
Are AI scribes used in New Zealand? Yes. The same international vendors operate in NZ. There is no public data on how many GP clinics use them, but adoption is growing. The lack of a NZ government warning does not mean the tools aren’t in use — it means the regulatory response is behind.
What’s the difference between an AI scribe and a human medical scribe? A human scribe is a trained professional in the room, bound by employment contracts and professional codes. An AI scribe sends audio to a third-party vendor’s cloud, processes it through a model, and generates a summary with no human oversight of the intermediate steps. The data handling risks are fundamentally different.
What should NZ regulators do? Issue a joint advisory from the Ministry of Health and the Privacy Commissioner covering consent (opt-in), data retention (minimum necessary), vendor transparency, and patient notification. Use the Australian investigation as the basis — the tools and the risks are the same.
🔍 THE BOTTOM LINE
Australia’s health department is worried enough about AI scribes to say so publicly. The same tools are in NZ clinics, with no equivalent warning. The technology is not the problem — unregulated clinical adoption is. NZ regulators have a choice: act now on the back of Australia’s investigation, or wait for the first breach and explain why they didn’t.
