Artificial intelligence and digital health

AI projects and ethics review

Digital health and artificial intelligence are rapidly transforming research, offering new opportunities but also complex ethical and regulatory considerations. As a researcher, you must ensure your study is designed, conducted and overseen in line with evolving guidance and appropriate ethical standards.

Determining whether your artificial intelligence or digital health project needs HREC review is an important first step. The answer depends on what your project involves.

There are two common scenarios:

You’re developing or testing an AI tool

If your AI tool is intended for a diagnostic, therapeutic or clinical purpose, it’s likely classified as a medical device under Australian law.

Any clinical trial of a medical device requires HREC review before it can proceed. So, you’ll need to determine whether your tool meets the medical device definition — see below.

You’re using AI as part of a research study

Even if the AI tool you’re using isn’t classified as a medical device, your research may still require HREC review if it involves human participants.

For example, using an algorithm to analyse participant health data, or deploying a large language model (LLM) to process interview responses, would typically require ethics review — regardless of whether the tool itself is regulated as a device.

If you’re not sure which scenario applies to you, contact us and we’ll help you work it out.

When AI meets the ‘medical device’ classification

Under Australian law, AI is classified as a medical device if it is intended for a therapeutic or diagnostic purpose, including:

• Diagnosis, screening, prevention, monitoring, prediction, prognosis, or treatment of a disease, injury or disability
• Alleviation of or compensation for an injury or disability
• Investigation of the anatomy or of a physiological process
• Control or support of conception.

AI that meets any of these criteria is regulated as a medical device, regardless of its form — including apps, websites, software programs, internet-based services or software packages. It may operate on a watch, phone, tablet, laptop or other device, be embedded in medical-device hardware, or function as part of a cloud-based ecosystem or standalone product.

Software incorporating generative AI — such as large language models (LLMs), text-generation systems or multimodal generative AI tools — is also regulated as a medical device if it performs a function that meets the above criteria.

AI products that do not diagnose, treat or inform clinical care are not classified as medical devices. These tools typically support social, behavioural or operational research without influencing health-related decisions.

For full guidance on classification, see the TGA Software and Artificial Intelligence Overview.

What Bellberry’s ethics committees can review

Bellberry’s HRECs are equipped to review:

• Clinical trials of AI tools classified as medical devices
• Research studies that use AI to collect, process or analyse data involving human participants
• Digital health studies where AI supports clinical decision-making or participant interaction

If your project involves AI in a purely operational or administrative capacity, with no participant interaction or health data, HREC review may not be required.

How to submit your AI study for review

When you submit an application involving AI, the HREC will consider the tool’s intended purpose, risk profile, data governance arrangements and participant impacts.

AI applications are submitted through Bellberry’s eProtocol online system.

See How to Apply for step-by-step guidance on registration, documentation requirements and the submission process and our AI Research Guidance document to help you meet the ethical review requirements for AI related studies.