AI scribes are conquering doctors' offices faster than they can be monitored

By The Guardian · July 4, 2026.

One figure captures the pace of change: according to a survey by the Royal Australian College of General Practitioners, the use of AI scribes—tools that record, transcribe and summarize the doctor-patient conversation to generate clinical notes—rose from 22% in August 2024 to 40% in November 2025. The companies selling these systems claim hundreds of millions of uses worldwide in a year and a half. Internal documents from Australia's federal health department, obtained by Guardian Australia via freedom of information laws, reveal the other side: the administration itself admits these tools 'have little oversight' and that some providers deliberately market themselves outside the regulatory radar, presenting themselves as not subject to medical device law or as 'privacy compatible' without real transparency about where they process the data. The department explicitly notes that some providers may not even know that their cloud platforms send patient data outside Australia.

There is also a detail that deserves attention: the government documents that some providers advertise to doctors a 30% increase in their income without working more hours or seeing more patients, which has direct implications for public spending on the medical reimbursement system (Medicare Benefits Scheme). In other words, the efficiency AI sells not only saves clinical time: it can also inflate billing without it being clear whether that translates into better care or simply more billable activity, as the director of the Consumer Health Forum, Elizabeth Deveny, warns. Added to this is a consent problem: the variation in how doctors ask patients for permission is enormous, and Guardian Australia has documented cases of psychiatrists turning away patients who do not accept the use of the scribe, or patients being told to seek another provider if they do not consent.

Australia's regulatory architecture illustrates a structural problem, not just a local one: oversight is split between the Therapeutic Goods Administration (which decides whether the scribe counts as a 'medical device'), the health practitioner accreditation agency (Ahpra) and the privacy commissioner. That division of responsibilities is precisely the gap that the most aggressive providers learn to exploit: if no one is clearly responsible, no one audits in depth. The privacy commissioner, Carly Kind, acknowledges that her office has spent months 'closely following' the deployment of the technology and that deficiencies persist in consent protocols and in the absence of any mention of scribes in clinics' privacy policies.

Our reading is that this episode is a textbook case of what happens when a technology with real, demonstrable benefits—reducing administrative burden and physician burnout, something the health department itself recognizes as a legitimate improvement—is deployed faster than the institutional capacity to put guardrails on it. It is not a problem of AI itself: transcribing and summarizing consultations is a task where language models perform well and free up valuable clinical time, aligned with that long-term promise of a more efficient, patient-centered healthcare system. The problem is the governance vacuum that opens up during the phase of mass adoption, when the commercial pressure to sign up clinics pushes some providers to blur their regulatory status and minimize transparency about where the most sensitive health data that exists travels.

This also marks a pattern we will see repeated: any healthcare system that adopts agentic AI in the doctor's office will face the same tension between speed of adoption, cost pressure (Medicare or other reimbursement systems) and regulatory fragmentation. Countries that resolve that governance sooner—clear classification of these tools, standardized informed consent, auditing of where data resides—will not only protect patients better, but will build the public trust needed for medical AI to scale without setbacks. The transition is uncomfortable, but the destination—a primary care less saturated with paperwork and with more human time for the patient—remains the right one if it is done with the appropriate safeguards.