Patients report being told they must find a different doctor if they refuse consent
Across Australia, a quiet revolution in clinical documentation is outpacing the institutions designed to govern it. AI scribes — software that records, transcribes, and summarizes patient consultations — have spread to two in five general practitioners in little more than a year, carried forward by promises of efficiency and relief from administrative burden. Yet the speed of adoption has exposed a familiar tension in technological progress: the tools arrive before the wisdom to wield them responsibly, leaving patients uncertain about what is being done with the most intimate details of their health.
- AI scribe adoption among Australian GPs nearly doubled in fifteen months, with patient conversations being recorded, transcribed, and processed — often by overseas servers — before any coherent regulatory framework existed to govern the practice.
- Federal health officials have identified a cascade of failures: vendors claiming exemption from oversight, unsubstantiated privacy compliance, and a fragmented three-agency regulatory structure that has left no single body clearly in charge.
- Patients are being forced to choose between accepting a technology they may not understand or losing access to their doctor entirely — a coercive dynamic that advocates say fundamentally undermines the meaning of informed consent.
- Regulators are now in motion: the privacy commissioner is monitoring implementation gaps, the Therapeutic Goods Administration is reviewing whether AI scribes qualify as medical devices, and the Consumer Health Forum is fielding a rising tide of complaints.
- The deeper unresolved question is whether the time AI scribes save clinicians is being returned to patients as better care, or quietly converted into higher Medicare billing volumes with no corresponding improvement in access or outcomes.
Australian general practitioners have embraced AI scribes with striking speed. The software sits quietly in the background of consultations, recording and transcribing conversations before generating clinical notes automatically. Where one in five GPs used the technology in mid-2024, two in five did so by late 2025 — a pace of adoption that has left regulators scrambling to catch up.
Internal federal health department documents, obtained by The Guardian, reveal the scale of the problem. Patient data is being transmitted to overseas servers without transparency. Vendors are marketing their products as falling outside regulatory requirements. There is no consistent standard for how — or whether — doctors seek patient consent. One claim circulating among practitioners promises a 30 percent revenue boost without additional hours, a figure that raises uncomfortable questions about where the efficiency gains are actually going.
Oversight is divided among three agencies — the Therapeutic Goods Administration, the health practitioner regulator, and the privacy commissioner — none of which has yet established clear authority. The TGA is currently deciding whether AI scribes should be classified as medical devices, a classification that would trigger formal scrutiny. That determination is expected soon.
The consent question is the most immediate pressure point. Patients who decline the technology have been told to find another provider. In Melbourne, a psychiatrist refused to continue treating those who would not agree to its use. The Consumer Health Forum has received growing numbers of complaints from patients who felt they had no real choice. Its chief executive, Elizabeth Deveny, argued that genuine informed consent requires genuine comprehension — and asked plainly whether time saved by AI is being reinvested in care or simply in billing.
Privacy Commissioner Carly Kind confirmed her office is actively monitoring the rollout, noting that many practices lack proper consent protocols and fail to disclose AI scribe use in their privacy policies. Officials acknowledged the technology's real benefits — reduced administrative load, potential relief from clinician burnout — while flagging the risks inherent in any large language model: errors, hallucinations, and inaccuracies that could compromise both patient safety and the integrity of medical records.
The technology is already embedded in thousands of practices. Whether oversight can close the gap between what patients believe is happening and what is actually occurring with their data now depends on how swiftly, and how seriously, regulators choose to act.
Australian doctors are increasingly turning to artificial intelligence to handle the paperwork of patient care. These AI scribes sit in the background during consultations, recording conversations, transcribing them, and automatically generating the clinical notes that doctors would otherwise have to write by hand. The technology has taken off with remarkable speed. In August 2024, one in five Australian general practitioners used an AI scribe. By November 2025, that number had climbed to two in five. Globally, companies selling the software report it has been deployed hundreds of millions of times in just the past eighteen months.
But as the tools have proliferated, Australia's federal health department has begun flagging serious problems. In internal briefing documents obtained by The Guardian, officials outlined a cascade of concerns: the technology operates with minimal regulatory oversight; some companies are sending patient data to servers outside Australia without making this transparent; and there is no consistent standard for how doctors ask patients for permission to use the tools. The department noted that some vendors explicitly market their products as falling outside regulatory requirements, or claim privacy compliance without substantiating it. One particularly troubling claim circulating among practitioners is that AI scribes can boost revenue by 30 percent without requiring additional hours or seeing more patients—a promise that raises questions about whether the time savings translate to better care or simply more billable activity.
The regulatory landscape is fragmented. Three separate agencies share responsibility for overseeing AI scribes: the Therapeutic Goods Administration, which regulates medical devices; the Australian Health Practitioner Regulation Agency, which oversees doctors themselves; and the Office of the Australian Information Commissioner, which handles privacy. The TGA is currently reviewing whether AI scribes should even be classified as medical devices, a determination that would trigger formal oversight. That review is expected to conclude in the coming months.
The consent problem is perhaps the most immediate. Patients report being told they must either accept the use of an AI scribe or find a different doctor. In Melbourne, a psychiatrist refused to continue treating patients who would not consent to the technology. The Consumer Health Forum, which advocates for patient interests, has received growing numbers of complaints from people who felt pressured into accepting the tools without understanding what they were agreeing to. Elizabeth Deveny, the forum's chief executive, emphasized that genuine informed consent requires patients to actually comprehend both the benefits and the limitations of the technology. She raised a harder question: if doctors are saving time with AI scribes, the public needs to know whether that time is being reinvested in better care and shorter wait times, or whether it simply means more appointments billed to Medicare.
Carly Kind, Australia's privacy commissioner, said her office has been closely monitoring the rollout. Her team has met with the Royal Australian College of General Practitioners, individual technology vendors, and civil society groups raising alarms. She highlighted specific gaps: many GPs are not implementing proper consent protocols, and few are disclosing the use of AI scribes in their privacy policies. The department acknowledged that the technology does offer genuine benefits—it can reduce administrative burden and potentially ease clinician burnout—but it also carries the same risks as any large language model: errors, hallucinations, and gaps in accuracy that could compromise patient safety and the integrity of medical records.
What happens next depends partly on the TGA's review and partly on how aggressively regulators choose to act. The health department has signaled it wants stronger, more consistent safeguards. But the technology is already embedded in thousands of practices across the country, and the momentum behind it is substantial. The question now is whether oversight can catch up before the gaps between what patients understand and what is actually happening to their data become too wide to close.
Citas Notables
Informed consent requires consumers to understand the benefits and limitations of the technology to which they are consenting— Federal health department briefing document
The public deserves to know if that time means better care, better access, or if it just means more billable activity, which doesn't help us with the burnout issue— Dr Elizabeth Deveny, Consumer Health Forum
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter if a doctor uses an AI scribe instead of typing notes themselves? Isn't it just a convenience tool?
It matters because the scribe is recording and transcribing the entire conversation between you and your doctor. That's sensitive information—your symptoms, your medical history, things you might not want stored anywhere. If that data is being sent to servers overseas without your knowledge, or if you're being pressured to consent without understanding what you're agreeing to, that's a privacy and autonomy issue.
But the technology is being used hundreds of millions of times globally. Surely it's safe if that many doctors are using it?
Scale doesn't equal safety. The problem is that there's almost no oversight. Some companies are explicitly marketing their products as falling outside regulatory requirements. Doctors are adopting the tools without consistent standards for how they ask patients for permission. And the health department found that some vendors don't even disclose that they're sending data outside Australia.
What's the concern about the revenue claim—the 30 percent increase without more hours?
It suggests the time savings aren't being used to see more patients or spend more time with each patient. If that's true, then the technology isn't solving the burnout problem or improving access to care. It's just making the same work faster, which means more billable activity per hour. That benefits the practice financially but doesn't help the actual problem doctors face.
Are patients actually being denied care if they refuse?
Yes. There are documented cases of doctors telling patients they need to find a different provider if they won't consent to the AI scribe. That's coercive. It puts patients in a position where they either accept something they don't fully understand or lose access to their doctor.
So what needs to happen?
Clearer rules about what counts as a medical device and who oversees it. Consistent, transparent consent processes. Disclosure in privacy policies. And regulators need to move faster than they currently are, because the technology is already embedded in thousands of practices.