Guide to TGA regulation of AI medical devices in Australia

AI medical devices: what Australian doctors need to understand about TGA regulation

Understanding whether a tool is regulated as an AI medical device can help doctors make more informed decisions about which tools are appropriate to use in their clinical practice.

As AI becomes more widely used in healthcare, it's also becoming more important to understand the rules and regulations around them. In February 2026, the Therapeutic Goods Administration (TGA) published updated guidance clarifying how AI-enabled software is regulated as a medical device in Australia.

When does a tool become an AI medical device?

Not every AI tool used in healthcare is regulated as a medical device. Instead, the TGA regulates according to what it is intended to do rather than the technology behind it, an approach known as technology agnostic regulation.

"The question isn't whether something uses AI," says Phil O'Sullivan, partner and Healthcare Sector Leader at law firm Allens. "The question is: what is its intended purpose, and what level of risk does that purpose create?"

If an AI tool is intended to diagnose, prevent, monitor, predict, provide a prognosis for, or treat it will generally fall within Australia's medical device framework. Examples of AI medical devices include:

AI used for administrative tasks such as scheduling, billing, drafting clinic letters or summarising consultation notes is unlikely to be regulated because it is not making clinical recommendations.

“You can have unregulated products, which are completely fine. Scheduling, billing and document tools, as well as pure transcription or note-taking functionality on its own, are unlikely to be classified as medical devices if they're not making clinical recommendations.”

However, O'Sullivan adds that the TGA is actively reviewing tools marketed as scribes that also generate diagnostic or treatment suggestions, so this is an evolving area of regulation that doctors should be aware of.

What the TGA's new AI medical device guidance means for doctors

As well as reinforcing that AI tools are classified according to their intended purpose, the updated guidance expands on how existing medical device regulations apply to AI-enabled software.

It includes guidance for manufacturers on managing software updates that change a product's intended purpose or performance, and sets expectations around the evidence needed to support AI-enabled medical devices.

"The regulation is really about regulating the supply of product, not so much an individual clinician user," says O'Sullivan.

However, clinicians still have an important role in deciding whether an AI tool is appropriate for the clinical task and how much weight to give its recommendations.

“Those decisions may have implications for patient safety, as well as professional liability, medical indemnity and reputation,” says O’Sullivan.

Choosing an AI medical device for clinical practice

When choosing an AI tool, clinicians should start with a few practical checks.

"Search the tool on the Australian Register of Therapeutic Goods (ARTG)," O’Sullivan says. "See if it's regulated and then compare that to what you're being told about the product in terms of what it's for, what it does, what the purpose is, and how it's being promoted."

Finding that a product isn't listed on the ARTG doesn't necessarily mean it shouldn't be used. Administrative AI tools may not require registration. But if a vendor claims its software can diagnose, predict or recommend treatment and it isn't registered, clinicians should start asking questions.

"Who is telling me that it can do these things? And are they standing behind that?" says O'Sullivan. "Ask about the evidence base. Is the tool clinically validated? What was the patient population? Was it trained on real data or synthetic data?"

As AI becomes more common in healthcare, O'Sullivan says responsibility can sometimes feel ‘a bit like that Spider-Man meme’, with vendors, clinicians and patients all pointing at each other.

"The vendor says, 'Exercise your clinical judgment, but my product's still good.',” says O’Sullivan. “The clinician says, 'If responsibility is still 100% with me, what's the benefit of providing the product?' And the patient probably looks at both of them and says, 'I don't really care as long as I get a good decision.’”

Whether or not an AI product is regulated, clinicians should think critically about what it's designed to do, the claims being made about it and if it's appropriate for the intended clinical task.

For doctors working within hospitals or health services, it is also recommended to check whether there are existing governance frameworks around AI. Many health services are now developing approved and prohibited use cases for AI tools, particularly where patient data is involved.

"If you're within a health network or institution, check not only the ARTG but whether there are organisational governance frameworks for the use of AI tools too," O’Sullivan says.

Never outsource clinical judgement

Selecting an appropriate AI tool is only the first step. Clinicians also need to think carefully about how they use it in practice.

"Your duty of care doesn't pause because you've used an AI tool, so always exercise 100% of your clinical judgment,” says O'Sullivan. “Don't rely on it. That's the safest way to do it, particularly if there's a chance the tool has produced an incorrect diagnosis or inappropriate treatment recommendation."

AHPRA makes a similar point in its guidance, stating that "TGA approval of a tool does not change a practitioner's responsibility to apply human oversight and judgment to their use of AI."

It also advises practitioners to make sure that any AI used is fit for purpose before introducing it into clinical practice and, where generative AI scribes are used, to check the accuracy and relevance of AI-generated records.

"I'm really optimistic about what well-designed, properly validated AI tools can do," O’Sullivan concludes. "They have enormous potential to improve diagnosis, reduce error and support clinicians who are under enormous pressure. But that potential comes with responsibility. The answer isn't to avoid these tools. It's to understand how to use them safely and appropriately."

AI medical devices: Key takeaways for doctors

What are the TGA guidelines for medical AI?

The TGA guidelines for medical AI explain when AI-enabled software is regulated as a medical device in Australia. Whether an AI tool falls within the framework depends on its intended purpose rather than the fact that it uses AI.

Are all AI tools used in healthcare regulated by the TGA?

No. AI used for administrative tasks, such as appointment scheduling or many transcription tools, is generally not regulated as a medical device. Regulation depends on the tool's intended purpose rather than the fact that it uses AI.

How can doctors check whether an AI tool is regulated?

If an AI tool is intended to function as a regulated medical device, clinicians can search the Australian Register of Therapeutic Goods (ARTG) and compare the product's intended purpose with how it is being promoted and used.

Brooke Nolan is a freelance journalist and writer specialising in health, medicine, science and technology. She also works as a strategic communications consultant, helping researchers, research institutes and not-for-profit organisations across Australia communicate evidence beyond academia.

Subscribe to the free InSight+ weekly newsletter here. It is available to all readers, not just registered medical practitioners.