TGA self-check

Is your AI a medical device?

The Therapeutic Goods Administration regulates software, including AI chatbots, symptom checkers and digital scribes, by its intended purpose, not by the technology behind it. If your product is presented as diagnosing, monitoring, predicting, treating or preventing a disease or condition, it may be a regulated medical device. Five questions, about two minutes, no account needed.

General information only

This self-check is an educational guide to the questions the TGA framework asks. It is not legal, regulatory or medical advice, and its result is not a determination of your product's status. Only the TGA can determine whether a product is a medical device, and borderline cases turn on details no questionnaire can capture. If your result is anything other than clearly out of scope, confirm your position with the TGA or a regulatory affairs professional before relying on it.

The self-check

Your answers stay in your browser. Nothing is stored or sent anywhere.

1.What is your software or AI product intended to do?

Think about what the product is for, as a user would understand it.

2.How do people rely on its output?

3.What do your website, app-store listing, instructions or sales material say it does?

The TGA reads intended purpose from what you publish, so this question matters as much as the product itself.

4.Which best describes what the AI itself produces?

5.Who controls the decision the software informs?

What the TGA actually looks at

Intended purpose

Whether software is a medical device turns on what it is intended to do: diagnose, prevent, monitor, predict, treat or alleviate disease. Intended purpose is read from your documentation, labelling, instructions and marketing, not from what your engineers had in mind.

The claims you publish

Clinical claims in a landing page, app-store listing or sales deck can bring a product into regulation on their own. General wellness products stay out of scope only while their claims stay general. Feature creep and marketing creep both count.

Evidence and registration

Regulated software must be included on the Australian Register of Therapeutic Goods before supply, classified by risk, and backed by evidence of safety and performance, including documented model design, training data and validation for AI.

Where certification fits

TGA regulation and Responsible AI certification answer different questions. The TGA regulates whether a product is safe and performs as claimed. Certification demonstrates that your organisation governs its AI responsibly: documented design decisions, training data quality, validation, risk management and human oversight. They stack. The governance evidence the TGA expects for AI devices is the same discipline certification assesses, whichever side of the line your product lands on.

To be clear: Responsible AI Australia is not a regulator, and certification is not TGA approval and does not substitute for it in any way.

Start your certification application Read the full explainer