Australia now has almost 960,000 registered health practitioners, and around 2.2 million people employed across health care and social assistance. AI tools are already in front of a large share of them. The question the sector has not answered is a simple one: who is responsible for teaching that workforce to use AI safely? Source and Source.
The Australian Alliance for Artificial Intelligence in Healthcare put that question near the centre of its third national agenda, “Don’t Walk. Run. AI in healthcare”, announced on 22 June 2026 and published in July 2026. Alongside it sits the Australian Commission on Safety and Quality in Health Care’s AI Clinical Use Guide, which already tells clinicians to educate themselves on how AI tools operate. This article looks at where responsibility actually sits, across every profession we assist. Source and Source.
Key Summary
- The Roadmap: “Don’t Walk. Run. AI in healthcare” sets out 8 priority areas and 28 recommendations, with a stated horizon of 2 to 5 years. (AAAiH)
- Workforce is Priority Area 5: understand workforce knowledge gaps and provide training in the use and implementation of AI-enabled healthcare services, covered by recommendations 16 to 18. (AAAiH)
- Scale of the task: almost 960,000 registered practitioners at 30 June 2025, and about 2.24 million people employed in health care and social assistance. (Ahpra, AIHW)
- Adoption is running ahead of education: 18.7 per cent of GPs reported routine use of AI scribes in May 2026, up from about 8 per cent in October 2024. (Healthed survey)
- The teachers are not trained either: 60 per cent of senior health educators reported they had never taught digital health. (ADHA)
- No mandate exists: Ahpra and the National Boards issued AI guidance in August 2024, but no National Board requires AI training and no body accredits AI education for clinicians. (Ahpra)
- The duty has already landed on the individual: the Commission’s guide asks clinicians to educate themselves, through their organisation or externally. (ACSQHC)
Adoption Has Outrun Education
Two numbers, two years apart, tell the story. In 2024, at least 80 per cent of GPs described themselves as not at all familiar or not very familiar with specific AI tools, and only 9 per cent felt familiar with AI-powered clinical decision support. Source.
By May 2026, a survey of 1,535 GPs found 18.7 per cent were using AI scribes routinely, more than double the roughly 8 per cent recorded in October 2024. Among those users, 37.3 per cent reported using scribes in 80 to 100 per cent of consultations. Source. Scribes have moved far enough into practice life that some employers now advertise them as a drawcard, which we covered in AI scribes as a recruitment perk.
The teaching side has not moved at the same pace. When the Australian Digital Health Agency released its Digital Health Train the Trainer Toolkit on 25 February 2026, it reported that 60 per cent of senior health educators had never taught digital health. Source.
The prize is not small, which is part of why adoption keeps moving. The Productivity Commission found in May 2024 that up to 30 per cent of tasks currently undertaken by the healthcare workforce could be automated using digital technology and artificial intelligence, and identified up to $5.4 billion a year in savings from reducing the time patients spend in hospital. Source.
Practical read: the tools are arriving through clinicians and practice managers, not through a training pathway. Capability is being built after the tool is already in the consulting room, if it is being built at all.
What the Roadmap Asks For on Workforce
“Don’t Walk. Run. AI in healthcare” was written by Professor Enrico Coiera and colleagues and published by Macquarie University in July 2026, drawing on consultation with 96 organisations. It contains 8 priority areas and 28 recommendations, with a stated implementation horizon of 2 to 5 years. Source.
Workforce is Priority Area 5. Its stated aim is to “understand workforce knowledge gaps and provide training in the use and implementation of AI-enabled healthcare services”. Three recommendations sit under it. Source.
- Recommendation 16: AI governance standards should encompass workforce training requirements, so healthcare organisations can deliver safe, secure and effective services using AI.
- Recommendation 17: safe, secure and effective use of AI should be a component of clinical training and digital health training programs.
- Recommendation 18: a national Chief AI Officer training program should be resourced, to produce a new generation of executive leaders and to upskill existing leaders such as CIOs.
The Roadmap is specific about who should deliver this. It points to Ahpra as a developer of professional standards well placed to drive workforce education, to professional colleges as the owners of medical trainee curricula, to the Commission as the body that sets standards informing health service training needs, and to universities and organisations such as the Australasian Institute of Digital Health as providers of targeted educational programs. Source.
That is a clear map of responsibility. The gap is what those bodies have actually built.
What Exists Today, and the Gap It Leaves
The Commission asks clinicians to educate themselves
The Commission published its AI Clinical Use Guide, Version 1.0, in August 2025. It is organised around three phases: before you use AI tools, while you use them, and after you use them. It sets out eight obligations for clinicians, and the seventh is the one that matters most here: “Educate yourself on how AI tools operate, either through your organisation or through external avenues.” Source.
The guide is clear about where accountability lands. Clinicians remain responsible for confirming the accuracy and completeness of records, results or documentation that will be acted on, and must always critically evaluate AI outputs, recognising that these tools support but do not replace clinical judgement. It also asks that records be labelled to show AI was involved in their creation, and that clinicians only use tools authorised or supplied by their organisation. Source.
The guide also notes, plainly, that evidence about the performance of AI tools in clinical settings may lag behind their implementation. It does not set a curriculum, a credential, or a required volume of training. Source.
The regulator restated existing obligations rather than creating new ones
On 22 August 2024, Ahpra and all National Boards published guidance titled “Meeting your professional obligations when using Artificial Intelligence in healthcare”. It covers accountability, understanding the tool, transparency with patients, informed consent, and legal and ethical compliance. It applies the existing codes of conduct to AI. It does not create an AI training requirement. Source.
College coverage is uneven
Some colleges have gone further than position statements. RANZCR has published ethical principles for AI in medicine, standards of practice for AI in radiation oncology, and papers on autonomous AI and generative AI, and runs AI content for trainees. Source. Imaging has also been the setting for Australia’s early publicly funded evaluation work, which we looked at in AI in radiology and what it means for imaging teams.
The RACGP position statement on artificial intelligence in primary care, updated in April 2025, states that GPs need instruction spanning medical informatics, mathematical concepts, data science and the ethics of AI, and commits the college to supporting GPs to develop those skills. Source. ACRRM published its own position statement in July 2024, holding that AI should support, not replace, clinical decision-making. Source. The RACP position statement goes furthest on education, arguing that physicians and trainees at all stages must be provided with the education and support required to use AI applications safely, effectively and responsibly. Source.
For pharmacy, dentistry, nursing and midwifery, the published position is largely the shared August 2024 board announcement. That is a wide variation in support for professions facing similar tools.
The national capability framework predates the problem
The Australian Digital Health Capability Framework, Version 1.2, was published in August 2024 by the Australasian Institute of Digital Health on behalf of the Australian Digital Health Agency. It covers the whole health workforce across five domains: digital professionalism, leadership and advocacy, data and information quality, information enabled care and services, and technology. It does not contain a dedicated AI capability domain. Source.
| Body | What exists today | What the Roadmap points to |
|---|---|---|
| Ahpra and the National Boards | Guidance from 22 August 2024 applying existing codes of conduct to AI. No training requirement. | Driving workforce education through professional standards and the common professional capabilities program. |
| Professional colleges | Uneven. RANZCR has principles, standards and trainee content. RACGP, RACP and ACRRM have position statements. Several professions have neither. | Ownership of trainee curricula, with safe AI use built into clinical training. |
| The Commission (ACSQHC) | AI Clinical Use Guide, August 2025, placing the education duty on the individual clinician. | Standards for healthcare organisations that inform operational and design training needs. |
| Universities and the AIDH | A digital health capability framework with no dedicated AI domain, and educators of whom 60 per cent had never taught digital health. | An educational marketplace supplying targeted AI programs. |
| Employers and health services | Local policies where they exist. The Commission asks clinicians to use only tools their organisation authorises or supplies. | Training requirements written into AI governance standards, and a resourced Chief AI Officer training program. |
What Practices, Health Services and Clinicians Can Do Now
None of the recommendations above will land quickly. The Roadmap works to a 2 to 5 year horizon, and the obligations in the Commission’s guide already apply. The practical work sits locally, and most of it is inexpensive.
1) Name someone accountable
- Give one person responsibility for AI tools in the practice or department, even part time.
- Keep a short register of which tools are approved, who uses them, and for what.
- Use an existing forum, such as a clinical governance or practice management meeting, to review performance rather than creating a new committee.
2) Restrict use to approved tools
- The Commission asks clinicians to use only AI tools authorised or supplied by their organisation. Make that list explicit rather than assumed. Source
- Confirm with the organisation or the developer where personal and sensitive information is stored and processed.
- Check whether the tool meets the definition of a medical device, and whether it is listed on the Australian Register of Therapeutic Goods.
3) Build verification into the workflow, not the policy folder
- Review AI-generated outputs during or at the completion of the patient interaction, while the detail is fresh.
- Label records to show AI was involved in their creation.
- Agree one consistent approach to patient disclosure and consent across the team, including what happens when a patient declines.
4) Induct new starters on the tools they will actually touch
- Cover the specific tools in use, their known limitations, and the local escalation path. Thirty minutes is better than nothing.
- Include automation bias: the tendency to prioritise an AI output over independent clinical judgement.
- Record that the induction happened. If capability becomes a governance standard, that record will matter.
Questions worth asking your employer
For clinicians assessing a role, AI capability is becoming a reasonable thing to ask about at interview. Four questions get to the substance quickly. For the wider picture on where AI now sits in hiring, see AI on both sides of the hiring desk.
- Which AI tools are in use here, and which are formally approved?
- What training or induction do you provide on them, and is it ongoing?
- Who is accountable when an AI output is wrong, and what is the escalation path?
- How do you handle patient consent and disclosure, and is the approach consistent across clinicians?
Why this matters. Accountability for AI outputs already sits with the clinician, whether or not training was provided. For employers, AI capability is moving from a compliance item to an attraction and retention factor, and candidates are starting to ask. For clinicians, the duty to educate yourself is already written down, so the safest position is to know the tools in front of you.
Conclusion
The Roadmap answers the headline question by naming the bodies that should teach the workforce: Ahpra, the colleges, the Commission, universities and the AIDH. What it also shows is that no single body owns the task today, and that the national frameworks were built before the tools arrived. Source.
Until that changes, capability is a local responsibility. The Commission’s guide is short, profession-agnostic and free, and it is a reasonable place for any practice or department to start. Source.
Disclaimer: This blog is a general overview and should not be construed as professional legal, financial or medical advice.
FAQs
- Is AI training mandatory for Australian health practitioners?
No. Ahpra and the National Boards published guidance in August 2024 applying existing codes of conduct to AI use, but no National Board mandates AI training and no body accredits AI education for clinicians. The 3rd National Policy Roadmap recommends that training requirements be built into AI governance standards.
- Who is responsible for training me to use AI safely?
In practice, you and your employer. The Commission’s AI Clinical Use Guide asks clinicians to educate themselves on how AI tools operate, either through their organisation or through external avenues. The Roadmap points to Ahpra, professional colleges, the Commission, universities and the AIDH as the bodies that should build the training system.
- Who is accountable if an AI output is wrong?
The clinician. The Commission’s guide states that you remain responsible for confirming the accuracy and completeness of records, results or documentation that will be acted on, and that AI tools support but do not replace clinical judgement.
- Does the 3rd National Policy Roadmap create new obligations?
No. It is a policy agenda, not a regulation. It sets out 8 priority areas and 28 recommendations for governments, regulators, colleges and industry, with a stated horizon of 2 to 5 years. Workforce sits at Priority Area 5, covered by recommendations 16 to 18.
- How many Australian clinicians are already using AI?
Comprehensive cross-profession data is limited. In general practice, a May 2026 survey of 1,535 GPs found 18.7 per cent were using AI scribes routinely, up from about 8 per cent in October 2024. Adoption in other professions is less well measured.
- What should a practice do first?
Name one accountable person, write down which AI tools are approved, and agree a consistent approach to reviewing outputs, labelling AI-assisted records, and patient consent. The Commission’s AI Clinical Use Guide is a short, free starting point that applies across professions.
Information Sources
- Australian Alliance for AI in Healthcare, Don’t Walk. Run. AI in Healthcare: The 3rd National Policy Roadmap
- Coiera E, Magrabi F, Silvera-Tawil D, Tyagi S and Chan A, Don’t Walk. Run. AI in healthcare, The 3rd National Policy Roadmap, Macquarie University, July 2026
- Australian Commission on Safety and Quality in Health Care, AI Clinical Use Guide
- Ahpra, Meeting your professional obligations when using Artificial Intelligence in healthcare
- Ahpra, Annual report highlights workforce growth and stronger safeguards in changing health sector
- Australian Institute of Health and Welfare, Health workforce
- Productivity Commission, Leveraging digital technology in healthcare
- Australian Digital Health Agency, Digital Health Train the Trainer Toolkit media release
- Australasian Institute of Digital Health, Australian Digital Health Capability Framework, Version 1.2
- RACGP, Artificial intelligence in primary care position statement
- RACGP, General Practice: Health of the Nation, Artificial intelligence
- RANZCR, Artificial Intelligence
- ACRRM, Artificial Intelligence in healthcare, College Position Statement
- RACP, Using artificial intelligence in clinical practice position statement
- The Medical Republic, How many GPs are actually using AI scribes?
- Gorilla Jobs, AI Scribes as a Recruitment Perk: The Caveats Before You Advertise
- Gorilla Jobs, AI on Both Sides of the Hiring Desk: Where the Line Now Sits
- Gorilla Jobs, AI in Radiology: 2025 Trial, What It Means for Imaging Teams
