Analysis capabilities push tools into new territory…
Australia’s medical regulator has warned that most AI-powered medical scribes it reviewed may require regulation as medical devices, warning that the technology is evolving from passive transcription into systems capable of analysing information and generating outputs that could influence clinical decisions.
The Therapeutic Goods Administration review looked at 12 medical scribes – including Heidi Health and iMedX, which have both been approved by Health New Zealand for use – and found eight have ‘functions likely to meet the threshold for a medical device’. The common factor was that the systems did more than simply capture and organise information. Instead, they analysed information and generated outputs that could be used to influence healthcare decisions.
“It appeared that some suppliers may not have previously considered medical device regulatory requirements.”
Two required additional information to determine their regulatory status. That left just two that the TGA says ‘appeared unlikely to be medical devices, based on the available information’. None of the devices assessed were identified as medical devices. The TGA didn’t provide any information on which of the 12 products assessed – Heidi Health, Lyrebird Health, Microsoft Dragon Medical One, Abby Health, mAIscribe, AskSam, VoiceBox Scribe, Medilit, VoiceboxMD, ConsultNote AI, Carepatron and iMedX – fell into each category.
It notes that while many of the products were originally developed to record and transcribe clinician-patient conversations, many products now perform a much broader range of tasks, including analysing information, determining what is clinically relevant or generating clinical conclusions, recommendations or other outputs that can influence diagnosis, monitoring, prognosis or treatment of a disease, injury, disability or condition.
In the UK, the NHS has had guidance since mid-2025 requiring scribes to have at least Class 1 medical device status approval.
A growing market
The scribe market has experienced rapid growth as healthcare organisations look for ways to reduce admin workloads and clinician burnout. The TGA assessment identified around 100 suppliers of digital scribes in Australia. Meanwhile a Royal Australian College of GPs poll in November 2025 found 40 percent of GPs were using the scribes in their general practice, though that number has been considered conservative.
Similar figures were touted for GPs in New Zealand in August 2025, and Health New Zealand has rolled out the AI scribes to 1,250 emergency doctors and frontline staff across all public emergency departments, following a successful Hawke’s Bay emergency department pilot using Heidi. (Heidi has committed to having regional infrastructure in New Zealand but says that depends on when infrastructure is available. It’s currently hosted in Australia.) The two-month pilot saw average documentation time cut from around 17 minutes per patient to just over four minutes. That enabled, on average, an extra patient per shift seen by each clinician, with resident medical officers and nurse practitioners in some cases seeing an extra three or four patients per shift.
Alongside Heidi and iMedX, Health NZ’s National AI and Algorithm Expert Advisory Group (NAIAEAG) has also endorsed T-Pro and IntelliTek for use in New Zealand. But while it has endorsed the offerings it also advises a ‘precautionary approach’ especially considering ‘risks to breach of privacy, inaccuracy of output, bias, lack of transparency and data sovereignty’.
Risk and regulations
Those risks have seen the scribe garner some controversy with experts on both sides of the Tasman issuing notes of caution.
In August, ABCnews reported that an AI scribe made up serious, but false, claims that a patient was taking illegal drugs. The patient wasn’t aware of the claims until they read a post-op letter her specialist sent to the GP after surgery suggesting psychedelic mushrooms may have been the reason for bleeding around the kidneys.
Australian AI expert Kobi Leins flagged her concerns last year after she was turned away from a clinic after declining a medical specialist’s request to use AI transcription software. Leins, whose roles include technical expert with the ISO and Standards Australia, had reviewed the AI transcription tool and said she ‘would not want my kids’ data anywhere near’ it, noting most had not had AI Impact Assessments and didn’t comply with AI management standards.
Digital Rights Watch is among those calling for the systems to be regulated as clinical devices and tested for bias and accuracy.
Andrew Cullen, senior research fellow at the University of Melbourne’s School of Computing and Information Systems, agrees and says the government must ensure Australians’ private data is properly safeguarded, through both legislation and bodies such as the AI Safety Institute. That includes forcing AI service providers to clearly disclose what data they have access to, how they use it and who can access it.
“More broadly, there is a need for rigorous independent testing to ensure the AI systems with which we interact are fair, reasonable and unbiased for all Australians. This should include testing whether these systems produce genuine productivity gains that are beneficial for patients,” Cullen says.
Industry bodies on both sides of the Tasman have issued advice on the use of the technology.
Future moves
The review also highlights a broader challenge facing AI regulation.
Across industries, vendors have commonly positioned generative AI systems as assistants designed to improve efficiency rather than replace human decision making. Yet as AI tools become more capable, the line between administrative support and operational influence is becoming increasingly blurred.
Healthcare provides one of the clearest examples of this shift. An AI system that simply records a conversation presents a different level of risk from one that extracts clinical findings, identifies potential issues and generates content that may shape subsequent decisions by healthcare professionals.
The TGA says in its assessment digital scribe developers and deployers were determining their products were administrative tools and not medical devices, and so did not include their product in the Australian Register of Therapeutic Goods.
“It appeared that some suppliers may not have previously considered medical device regulatory requirements. The findings highlighted the need for clearer guidance on when digital scribes are regulated as medical devices and how suppliers can meet their obligations.”
Others included statements that they were not a medical device, or they were not intended for diagnosis, treatment or other medical purposes, but the TGA warns that statements or disclaimers ‘are not sufficient in determining if a product is, or is not, a medical device’.
It says it will use the review findings to inform its regulatory approach to digital scribes, and to guide targeted compliance activities where digital scribe functionality meets the definition of a medical device.
