AI Scribes in Australian Clinics: What AHPRA's Guidance Requires
Published 30 September 2026
Short answer: AHPRA's guidance on AI keeps practitioners personally responsible for how AI is used in their practice. For AI scribes that use generative AI, informed patient consent will generally be required, ideally noted in the health record, and practitioners must check the accuracy of the notes the scribe produces. Separately, on 3 August 2026 the TGA said its review of digital scribes has moved into compliance action against suppliers.
AI scribes have become close to standard in Australian clinics over the past couple of years, quietly listening to a consult and turning it into structured notes. What hasn't kept pace, in a lot of practices, is a clear answer to a simple patient question: where does my consult go?
What AHPRA's guidance actually asks of practitioners
AHPRA's guidelines, Meeting your professional obligations when using Artificial Intelligence in healthcare, apply the existing National Board codes of conduct to AI. They don't create a new rulebook. They make clear how the existing one applies.
- You remain accountable. Practitioners stay responsible for the care they deliver, whatever tool helped produce the record.
- Understand the tool. Practitioners need to understand enough about an AI tool to use it safely, including its intended use and limits. AHPRA specifically points to understanding how data is used to retrain the AI, where data is located, and how it is stored.
- Be open with patients and get informed consent. Practitioners should tell patients about their use of AI. For a generative AI scribe, AHPRA says informed consent will generally be required, ideally noted in the health record, and the software should include an explicit consent step before recording starts.
- Check the output. If you use an AI scribe, you're responsible for checking the accuracy and relevance of the records it creates. TGA approval doesn't change that.
Two developments that sharpen the question
The TGA has moved from education to enforcement. On 3 August 2026, the TGA said its year-long review of digital scribes has moved into compliance action against suppliers whose scribes operate as medical devices without registration. Not every scribe is a medical device, since it depends on the product's intended purpose, and this is a vendor obligation rather than a new duty on your practice. But it's a fair question to put to any scribe supplier: is this product a regulated device, and if so, is it on the ARTG?
Privacy law is heading the same way. The exposure draft Privacy Amendment (Personal Data Protection) Bill 2026, released for consultation on 31 August 2026, would define disclosure broadly: making personal information accessible to another person or body. It would also introduce a controller and processor framework, under which a practice can still be held responsible for what a vendor does on its behalf. It's a draft, with no commencement date, but it points the same direction as everything above.
What "informed consent" needs to actually cover
For consent to mean something, a patient needs to understand, at minimum:
- That an AI tool is being used to process what they say in the consult
- What happens to the audio and the resulting notes afterwards
- Where that information is stored, and whether it leaves the practice
- That they can decline
That third point is where a lot of practices get uncomfortable, because the honest answer is often "a cloud server run by our scribe vendor, and we're not entirely sure of the retention terms." That's not an easy sentence to say to a patient who has just disclosed something sensitive.
The alternative
Running AI scribe and clinical documentation tools on hardware inside the practice changes that conversation. The audio and notes never leave the building, so the consent conversation becomes concrete instead of vague: "your consult notes are processed on our own equipment, in this practice, and never sent to an outside provider."
Your review obligations stay exactly the same. AHPRA still expects you to check what the AI produces. What changes is what you have to explain to a patient.