Privacy policies need an automated decision-making section by 10 December 2026.
Draft yours free

AI in Australian healthcare

Healthcare is Australia's most regulated setting for AI. The TGA treats AI software as a medical device whenever its intended purpose is clinical, Ahpra holds practitioners personally accountable for AI outputs, and health information attracts the strictest privacy rules in the Privacy Act. Reforms flagged by the TGA's AI review are in progress, and AI scribes are an active compliance focus.

TGAAhpra / National BoardsOAICDept of HealthACSQHC

Last reviewed October 2026. Every entry links to its official source.

The rules

What applies today

The TGA regulates AI-enabled software as a medical device under the Therapeutic Goods Act whenever its intended purpose is diagnosis, prevention, monitoring, prediction, prognosis or treatment. The framework is technology agnostic: intended purpose, not the presence of AI, triggers regulation, and regulated devices must be in the ARTG before supply. The guidance covers scope creep (an update adding clinical features can turn a non-device into a device), off-label use, evidence expectations for training data and limits on synthetic validation data. On 8 September 2026 the TGA amended the clinical decision support system exemption to clarify when it applies, effective 1 November 2026; the TGA says the scope of the exemption is unchanged, so developers should re-read their exemption reasoning against the amended wording before November.

Who this affects: Developers and suppliers of AI software with a clinical purpose, including LLM chatbots and clinical decision support.

Latest updates

What changed recently

  1. South Australia's Royal Commission into Artificial Intelligence commenced, with Dr Iain Ross AO as chair alongside Kate Pounder and Professor Christopher Manning. It reports by 1 July 2027.

    ABC News: three experts appointed to lead SA's Royal Commission into AI

  2. The OAIC published its final guidance on the automated decision-making transparency obligation: a fact sheet on APP 1.7 to 1.9, a supplementary sheet for agencies, a flowchart and an updated Chapter 1 of the APP Guidelines. Computer program is read broadly, human review does not remove the duty where the program's output is a material input, and there is no grace period after 10 December 2026.

    OAIC: new resources on transparency for AI and automated decision-making

  3. The Office of AI in PM&C opened consultation on the Australian Standards for AI, proposing mandatory standards for large data centres and conditions for AI training rather than obligations on organisations deploying AI; submissions close 9 October 2026. Health services deploying AI are not caught.

    PM&C: have your say on AI training and infrastructure

  4. The TGA amended the clinical decision support system exemption to clarify when it applies, effective 1 November 2026, without changing its scope.

    TGA news

  5. Submissions closed on the TGA's consultation on the Therapeutic Goods Advertising Code, which proposes ARTG number display in software-as-a-medical-device advertising, software-specific mandatory statements and disclosure requirements for SaMD that incorporates AI. Outcome not yet published.

    King & Wood Mallesons: TGA consultation on the Advertising Code

  6. The Attorney-General released the Privacy Amendment (Personal Data Protection) Bill 2026 exposure draft, with a fair and reasonable test for handling personal information and express coverage of AI-derived information, relevant to health information handling; submissions closed 18 September 2026.

    AGD consultation: privacy reform

  7. National Cabinet agreed the Commonwealth will legislate nationally consistent AI laws and mandatory standards for large data centres in early 2027, complementing state and territory planning processes.

    National Cabinet communique

  8. Parliament appointed a Joint Select Committee on Artificial Intelligence, naming health and medical research among the sectors it will examine, reporting by 30 November 2026.

    Parliament of Australia: Joint Select Committee on AI

  9. South Australia announced Australia's first Royal Commission into Artificial Intelligence, commencing 1 October 2026 and reporting by 1 July 2027.

    SA DPC announcement

  10. The Australian Alliance for AI in Healthcare launched its third National Policy Roadmap for AI in healthcare, covering governance, workforce capability, regulation and AI safety. An industry and academic advocacy document rather than a regulator instrument.

    AAAiH: 3rd National Policy Roadmap

  11. The National AI Centre published updated implementation guidance for its Guidance for AI adoption, the voluntary six-practices governance framework that evolved the Voluntary AI Safety Standard.

    Guidance for AI adoption (industry.gov.au)

  12. The TGA published updated guidance on AI and medical device software regulation, clarifying when AI products including LLM chatbots and clinical decision support are medical devices, and how scope creep and off-label use trigger regulation.

    TGA AI guidance

  13. The TGA updated its digital scribes guidance: transcription-only scribes are not devices, but scribes adding diagnostic or treatment suggestions are, and must be in the ARTG before supply.

    TGA digital scribes guidance

Getting ready

Compliance checklist

  • Classify every AI tool against the medical device definition by its intended purpose, and confirm ARTG inclusion before deployment if it qualifies. See the rule

  • Re-assess after every software update. A new feature, like a scribe that starts suggesting diagnoses, can change intended purpose and require TGA approval before rollout. See the rule

  • Keep practitioners accountable and in the loop: human judgment applied to every AI output, and tools tested for fitness for purpose before clinical use. See the rule

  • Obtain and record informed patient consent where AI tools take patient data as input, especially consultation-recording scribes. See the rule

  • Do privacy due diligence before adopting any AI product: no health information in public AI tools, a valid basis for secondary use, and a privacy policy that discloses AI use. See the rule

  • Stand up organisational AI governance: an accountable owner, a risk treatment plan, pre-deployment testing and incident monitoring including notifiable data breaches. See the rule

  • Equip clinicians with the ACSQHC AI Clinical Use Guide and its ambient-scribe safety scenario, and check professional indemnity insurance covers AI use. See the rule

Straight answers

Questions people ask

When does the TGA treat AI software as a medical device?

Whenever the software's intended purpose is diagnosis, prevention, monitoring, prediction, prognosis or treatment. The test is intended purpose, not the presence of AI. Regulated devices must be included in the Australian Register of Therapeutic Goods (ARTG) before supply, and an update that adds a clinical feature can turn a non-device into a device. The TGA amended the clinical decision support system exemption on 8 September 2026, effective 1 November 2026.

Are AI scribes regulated in Australia?

Transcription-only AI scribes are not medical devices. A scribe that analyses or interprets clinical conversations, for example by generating a diagnosis or treatment recommendation the practitioner did not state, is a medical device and must be in the ARTG before supply. Practitioners using scribes must obtain informed consent, verify record accuracy and recheck ARTG status after software updates. The TGA has named AI and software-based medical devices as a compliance focus.

What are Ahpra's rules for health practitioners using AI?

Ahpra's guidance applies the existing codes of conduct to AI. The practitioner remains accountable for safe care and must apply human judgment to any AI output; TGA approval of a tool does not change that. Practitioners must understand the tool's training data and limitations, be transparent with patients and obtain informed consent, especially for scribes that record consultations. The guidance also covers algorithmic bias, employer governance and professional indemnity cover.

Can a clinic put patient information into ChatGPT?

The OAIC's position is that personal or health information should not be entered into publicly available chatbots. Health information is sensitive information under the Privacy Act, which applies to health services of any size. Feeding personal information into an AI tool engages the purpose-limitation rules, AI-generated or inferred information counts as a collection, and the privacy policy must disclose AI use.

Is there a national safety guide for using AI in patient care?

Yes. The Australian Commission on Safety and Quality in Health Care publishes the AI Clinical Use Guide, structured around checks before, while and after using an AI tool, with safety scenarios on ambient scribes and interpreting medical images. The TGA's digital scribes guidance cross-refers to it.

Is AI regulation in Australian healthcare changing?

The TGA's review of whether therapeutic goods legislation is fit for AI found the framework broadly sound but flagged refinements across definitions, classification rules, essential principles and excluded software, and government approved further work on its fourteen findings. No legislative amendment is in force yet. The Joint Select Committee on AI reports by 30 November 2026 and South Australia's Royal Commission into AI, which commenced on 1 October 2026, includes public health services in scope.

Show your customers you're across all of this

Certification against Australia's AI Ethics Principles is independent proof that your business uses AI responsibly, before a client, regulator or tender asks.

Get certified