Skip to content
Ignis
AI, Plainly

Client data and AI: a Privacy Act checklist for small Australian firms

Before an accounting practice, law firm or clinic puts client information through an AI tool, there are a handful of things worth checking. Here they are in plain English, including the automated decision-making rules that start on 10 December 2026.

Tav Palapathwala, Founder

7 min read

An open filing cabinet drawer full of index cards
Photo by Maksym Kaharlytskyi on Unsplash
On this page

The common worry about AI and client data is that the AI will do something strange with it. The bigger risk is more ordinary. Someone pastes a client's tax return, a file note or a referral letter into a free chatbot to save some time, and the firm no longer controls where that information went.

The good news is that the questions to ask are not new. They are the same ones the Privacy Act has asked for years, applied to a new kind of tool. Here they are in order, with the change that starts on 10 December 2026.

This is general information, not legal advice. Your obligations depend on your firm, your profession and your clients. If you are unsure, speak to your lawyer or your professional body.

First, check whether the Privacy Act covers you

Many small businesses with annual turnover of $3 million or less are not covered by the Privacy Act. Some are covered regardless of size, including private sector health service providers. The OAIC's small business guidance sets out who is in and who is out.

Even if your firm is exempt, the Australian Privacy Principles (APPs) are a sensible standard. Your clients expect it, larger clients may require it in contracts, and accountants and lawyers carry professional confidentiality duties whether or not the Act applies.

Do not put client information into public chatbots

This is the clearest line, and the regulators have drawn it.

The OAIC's guidance on privacy and the use of commercially available AI products, published in October 2024, recommends as best practice that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools.

For lawyers, the December 2024 joint statement from the Victorian Legal Services Board and Commissioner, the Law Society of NSW and the Legal Practice Board of WA is direct: "Lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots/copilots (like ChatGPT), or any other public tools."

For clinics, Ahpra's guidance on meeting your professional obligations when using AI in healthcare says practitioners should involve patients in decisions to use AI tools that require their personal data, and should understand whether patient data is used to train the tool.

The practical step is a written rule, shared with every staff member: which AI tools are approved for client work, and that everything else is off limits for client information.

The checklist before you use an AI tool on client data

The OAIC publishes two short checklists alongside its guidance, one for selecting a product and one for using it. The questions below follow the same ground in plain terms.

1. What exactly goes in? List the kinds of information the tool will see: names, tax file numbers, bank details, health information, matter details. Sensitive information, such as health information, carries stricter rules. Send the minimum the job needs.

2. Is the use consistent with why you collected it? Under APP 6, personal information is generally used for the purpose it was collected for, or a related purpose the client would reasonably expect. Using a client's documents to prepare their tax return is one thing. Using them to train a vendor's model is another.

3. Does the vendor use your data for training, and can you turn it off? Read the business terms, not the marketing page. Look for a clear statement on training, retention periods and who at the vendor can access the data.

4. Where is the data stored and processed? APP 8 deals with sending personal information overseas. You can use overseas providers, but you generally remain accountable for how they handle it. Ask the vendor which countries the data is stored and processed in, and get it in writing.

5. Is it secure enough? APP 11 requires reasonable steps to protect personal information. Check access controls, multi-factor sign-in, encryption, and what happens to data when you leave the service.

6. Is the output accurate enough to rely on? APP 10 is about keeping personal information accurate. AI tools can produce confident errors. Decide who reviews outputs before they reach a client, a regulator or a file.

7. Is there a record? Keep a log of what the tool did: what went in, what came out, who approved it and when. If a client asks, or something goes wrong, you can see what happened.

8. Do your privacy policy and notices say so? APP 1 requires a clear, current privacy policy, and APP 5 covers telling people about collection. If AI tools are part of how you handle client information, your policy and engagement letters should reflect that.

The new rule from 10 December 2026: automated decisions

The Privacy and Other Legislation Amendment Act 2024 added new APPs 1.7 to 1.9. From 10 December 2026, organisations covered by the Act must say in their privacy policy when they use computer programs to make decisions, or do things substantially and directly related to making decisions, that could significantly affect the rights or interests of an individual.

The policy must describe the kinds of personal information used in those programs and the kinds of decisions involved. The OAIC released its updated APP 1 guidelines and a fact sheet on this obligation in September 2026. As law firm summaries of that guidance note, a person reviewing the output does not by itself take a tool out of scope.

For most small firms, AI used to draft emails or extract data from documents is unlikely to be making decisions that significantly affect clients. But a tool that scores which clients get offered a payment plan, triages patient referrals or screens new matters might be. It is worth listing every place software makes or shapes a decision about a person, and checking each against the guidance before 10 December.

Human review is the safeguard that matters most

Most of this comes back to one design choice. The AI drafts, sorts and flags. A person who knows the client reviews and approves anything that goes out, gets lodged or gets paid.

That is not just good privacy practice. It is how professional obligations already work. Ahpra's guidance says the practitioner remains responsible for safe care whatever tools they use, and the legal regulators' statement makes the same point about lawyers' judgement. An automation that removes the person removes the protection.

It is also how we build every workflow: the system handles the boring work inside your own tools, and a person signs off on what matters. The steps are on how it works.

A short path for a small firm

Write the rule. One page: approved tools, what must never go into anything else, and who to ask.

List where client data meets AI today. Include the tools staff are using informally. That list is usually longer than the partners expect.

Check each approved tool against the eight questions above. Keep the answers on file.

Update your privacy policy and engagement letters, including any automated decision disclosures needed from 10 December 2026.

Review it every six months. Vendors change their terms and features often.

If you are thinking about automating document work, our guide to getting receipts and supplier invoices into Xero shows what a reviewed, exceptions-based workflow looks like. For phone work, see AI phone answering for trades and garages, which covers call-recording consent.

FAQ

Can my staff use ChatGPT or similar tools for client work?

The OAIC recommends not entering personal information, especially sensitive information, into publicly available generative AI tools, and the legal regulators say lawyers cannot safely put confidential client information into public chatbots. Business versions with appropriate contractual terms are a different question, and need checking against your obligations.

Does the new automated decision-making rule apply to my small firm?

It applies to organisations covered by the Privacy Act, from 10 December 2026. Many small businesses with turnover of $3 million or less are exempt, but health service providers and some others are covered regardless. Check the OAIC's guidance or ask your adviser.

Is it a problem if our AI provider stores data overseas?

Not automatically. APP 8 allows overseas disclosure in some circumstances, but your firm generally stays accountable for how the provider handles the information. Know where the data goes and make sure the contract covers it.

Do we need to tell clients we use AI?

If AI is part of how you handle their information, your privacy policy and collection notices should reflect it, and from 10 December 2026 some automated decisions must be disclosed. Clinics should also follow Ahpra's guidance on involving patients in decisions to use AI with their data.

Book my workflow audit

Ignis AI builds and runs AI automations for Australian businesses. People approve every output that matters.