What is an AI scribe in general practice?

An AI clinical scribe is a software tool that listens to a GP consultation and automatically generates clinical notes or a medical record entry. Many Australian general practices began using these tools from 2023 onwards to reduce administrative burden. The tools vary widely — some use cloud-based processing, others run locally — and their accuracy, transparency, and data handling practices differ significantly between products.

Is my GP required to tell me they are using an AI scribe?

Australian guidelines from the RACGP and the Office of the Australian Information Commissioner indicate that patients should be informed when an AI tool is used to record or document their consultation, as this involves processing of sensitive health information. In practice, disclosure has been inconsistent. If you are concerned, you can ask your GP directly before the consultation begins.

What does 'scope creep' mean in this context?

The TGA's review found some AI scribes were influencing clinical decisions beyond their intended function — for example, suggesting diagnoses, flagging medication interactions, or shaping what information appeared prominently in the clinical note. These functions take the tool from documentation assistant into territory that may constitute a regulated medical device under Australian law.

TGA moves toward enforcement on AI scribes after safety review

What just happened

At the Health Informatics Conference 2026 (HIC2026) this week, the Therapeutic Goods Administration’s head of product quality, Tracey Duffy, announced that the TGA is moving from educational engagement to enforcement against AI clinical scribe suppliers that have deployed these tools as medical devices without regulatory approval.

The announcement follows a year-long TGA review — the first systematic examination of AI clinical scribes in Australia. The findings are uncomfortable reading for any practice that has adopted these tools without scrutinising them closely.

The safety concerns the review identified:

Scope creep — AI systems influencing clinical decisions beyond their intended purpose
Lack of transparency and post-market feedback between developers and clinical users
Inadequate monitoring controls and risk mitigation procedures
Insufficient lifecycle management once systems are deployed in practice

Duffy indicated that enforcement actions are likely within 12 months, and that the TGA will publish detailed review outcomes and updated developer guidance “in the coming weeks.”

The context is not hypothetical. Digital rights and privacy advocates have begun scrutinising GP AI scribe tools — raising questions about where consultation audio is processed, how long it is retained, what happens to the data, and whether patients are meaningfully consenting to being recorded by a third-party AI system during some of the most private conversations they are likely to have.

Both-and

The case for AI scribes is real

Let’s be direct about what these tools are solving. Clinical documentation burden is one of the most significant contributors to GP burnout in Australia. The cognitive switch between being present with a patient and then producing accurate, legible, medico-legally defensible notes is significant — and it compounds across a full day’s appointments.

AI scribes that work well genuinely reduce that load. GPs who use them report being more present in consultations, less exhausted at the end of the day, and producing more detailed records. That is not a trivial benefit. The population-level value of keeping GPs in the workforce and mitigating burnout is a legitimate counterweight to the risks.

The tools also have real potential for improving continuity of care. Detailed AI-generated notes may capture more nuance than abbreviated notes written in the thirty seconds between patients.

The concerns are also real

The TGA’s finding of “scope creep” deserves careful attention. A scribe that listens and transcribes is one category of tool. A scribe that is subtly shaping the clinical record — determining what gets emphasised, flagging diagnoses, surfacing medication alerts — is functioning as a clinical decision support tool. That carries different regulatory expectations and a different risk profile.

The review found inadequate post-market monitoring. That means general practices adopting these tools have largely been doing so without a formal feedback loop to developers about errors, omissions, or safety events — and without developers having clear obligations to track outcomes at a population level.

The privacy dimension compounds this. A patient presenting with a mental health concern, a sensitive reproductive health question, or a workplace injury matter has not consented — in most cases — to that conversation being transmitted to a third-party cloud service for AI processing. The legal and ethical framework around this is still catching up to the pace of adoption.

The regulatory response is measured, not alarmist

The TGA’s review concluded that Australia’s existing legislative framework remains adequate — refinements are needed, but not new laws. The pathway forward involves strengthening post-market monitoring as a precondition for streamlining regulatory approval processes. Compliance action against specific non-compliant suppliers — rather than a blanket prohibition — is a proportionate response.

But “within 12 months” is a long runway. Between now and enforcement, practices will continue adopting and using tools whose regulatory status is unclear.

My two cents

The documentation burden that AI scribes address is genuine — I understand why colleagues have adopted them quickly and why the time savings become indispensable once they are embedded in workflow.

What the TGA review is articulating is something I think many GP users of these tools have quietly noticed: these systems are not neutral transcribers. They make choices. They shape what the clinical record looks like. When an AI scribe determines how to summarise a patient’s account of their pain, or whether a medication concern appears in the note, that is a clinical act — and clinical acts should be subject to clinical accountability, not just a software licensing agreement buried in an onboarding email.

The patient consent question is the one that matters most to me. When someone comes in distressed and needs to talk, the knowledge that a cloud system is processing that conversation changes the nature of the encounter. Informed consent is not just a legal formality; it is the foundation of the therapeutic relationship. Most practices using AI scribes are not yet having that conversation explicitly with patients.

If your practice uses an AI scribe, this week is a reasonable moment to ask: does the tool hold TGA clearance? What is the data handling policy? Are patients being informed and given the opportunity to decline recording?

Verdict: yes — worth knowing about, whether you are a patient attending a general practice or a GP using one of these tools.

Sources cited

TGA flags compliance action on AI scribes as review finds safety concerns. Medical Republic, 4 August 2026. https://www.medicalrepublic.com.au/tga-flags-compliance-action-on-ai-scribes-as-review-finds-safety-concerns/127888
Digital rights activists eye GP AI scribes. Medical Republic, 4 August 2026. https://www.medicalrepublic.com.au/digital-rights-activists-eye-gp-ai-scribes/127885