Z&Z
0403 089 563Call usBook a free chat

Privacy and safety · 8 min read

Is it safe to use ChatGPT with client information?

Levi Zimmerman, smiling, in a light-blue shirt

By Levi Zimmerman

Published · Updated

Short answer: it can be, but only on the right plan, with the right settings, and with clear rules about what goes in. The safest habit is simple: don’t put anything into ChatGPT that identifies a client or patient unless you’ve checked that your plan and settings protect it, and your obligations allow it.

It’s one of the first questions dental practice managers, lawyers, accountants and allied health teams ask us about AI at work. Here’s the longer answer.

This is general information, not legal advice. If you handle sensitive information, such as health records or legal matters, check your obligations with your adviser or professional body.

Why it’s a fair question

When you type into an AI tool, your words go to the provider’s servers so the AI can answer. What happens next depends on the tool and the account you’re using:

  • Is the conversation kept, and for how long?
  • Can people at the provider see it?
  • Is it used to train future versions of the AI?
  • Where in the world is it stored?

Those answers are different for a free personal account and a paid business account, even from the same company.

Personal accounts and business accounts are not the same

OpenAI, the company behind ChatGPT, treats personal and business accounts differently.

On personal plans (Free, Plus and Pro), OpenAI says your conversations may be used to train its models unless you switch that off. The setting lives under Settings, then Data Controls, and is called “Improve the model for everyone”. Turning it off stops your new conversations being used for training.

On business plans (the workplace and enterprise plans, and the API that software developers use), OpenAI says it doesn’t use your business data to train its models by default. Business plans also give an owner or administrator control over who’s on the account and how it’s set up.

Other tools work the same way. Microsoft Copilot and Google’s Gemini each have consumer versions and business versions inside Microsoft 365 and Google Workspace, with different protections. Whatever you use, check which version your team is actually logged into. A staff member using their own free account on a work laptop is an easy gap to miss, and a common one: Microsoft’s 2024 Work Trend Index found 84% of Australian knowledge workers use generative AI at work, and 78% of Australian AI users bring their own AI tools (Microsoft, 2024).

Terms and settings change, so check the provider’s current data controls page rather than relying on what was true last year.

What the privacy regulator says

The Office of the Australian Information Commissioner (OAIC) has published guidance on privacy and using commercially available AI products. Two points from it matter for most small businesses:

  1. Personal information can generally only be used for the purpose you collected it for, unless the person has agreed or would reasonably expect the other use. This comes from Australian Privacy Principle 6. Pasting a client’s details into an AI tool is a use of that information.
  2. As a matter of best practice, the OAIC recommends not entering personal information, and particularly sensitive information, into publicly available AI tools.

“But I’m a small business. Does the Privacy Act even apply to me?”

Many small businesses with an annual turnover of $3 million or less are exempt from the Privacy Act. But there are exceptions. For example, health service providers are covered regardless of size, and so are businesses that trade in personal information. Check the OAIC’s page on small businesses to see where you stand.

Even if you’re exempt, two things still apply:

  • Professional duties. Health practitioners, lawyers and accountants have their own confidentiality rules, and those don’t depend on turnover.
  • Trust. Your clients assume you’re looking after their information. “We were technically exempt” won’t help much if something goes wrong.

Seven habits that make it safe enough to use

This is the setup we put in place when we train teams.

1. Pick one approved tool, on a business plan

Choose one AI tool for the business, ideally the one that fits the software you already use, and pay for the business version. One tool is easier to secure, easier to train people on and easier to keep track of.

2. Check the settings before anyone uses it

Confirm that training on your data is off, that only your team can access the account, and how long conversations are kept. If you’re on a personal plan for now, at least switch off “Improve the model for everyone”.

3. Take out the names

Most of the time, AI doesn’t need to know who the client is to help. “Explain what a crown involves to a nervous patient, in everyday words” works just as well without the patient’s name, date of birth or health fund number. “Write a polite follow-up to a client who hasn’t paid an invoice from June” works just as well without the client’s name, address and account number. Swap real names for “Client A” and leave out identifying details.

4. Know what never goes in

Write a short list your whole team can remember. For most businesses it includes:

  • health information
  • tax file numbers, licence numbers and passport numbers
  • bank and card details
  • passwords and login details
  • anything a client has told you in confidence

5. Use temporary chats for one-off questions

ChatGPT has a Temporary Chat option. OpenAI says these chats don’t appear in your history and aren’t used to train its models, though it may keep a copy for a short time for safety checks. It’s handy for a quick question, but it doesn’t replace habits 3 and 4.

6. Check everything before it goes out

AI can sound confident and still be wrong. Anything going to a client, such as advice, figures or dates, gets read by a person first. Make that a rule, not a suggestion.

7. Write it down in one page

A one-page AI policy covers which tool is approved, what never goes in, who checks the output and who to ask when unsure. It turns good intentions into something a new starter can follow on day one.

Many workplaces don’t have one yet. In KPMG and the University of Melbourne’s 2025 study, only 30% of Australian employees said their organisation has a generative AI policy, and 48% admitted using AI in ways that go against workplace policy (KPMG and University of Melbourne, 2025). A clear page your team actually reads is where closing that gap starts. If you update your privacy policy to mention the AI tools you use, even better.

What to do this week

If you only do three things:

  1. Find out what your team is already using. Ask, without blame. Going by the figures above, chances are they already use something.
  2. Move everyone onto one business account, with training switched off.
  3. Share the “never goes in” list and talk it through at your next team meeting.

That gets you most of the way. The rest is training people to use the tool well, so they’re not tempted to cut corners when they’re busy.

Where we can help

This is exactly what our AI training covers. We help you choose the right tool and plan, set it up right, write your one-page policy with you, and teach your team the prompts and processes that get good results without putting client or patient information at risk.

Once the basics are safe, we can set up an AI assistant that knows your templates and house style. The step up is an AI brain: AI connected to your documents, systems and processes, answering your team’s questions and powering your automations from one place.

Our first conversation is free. If you’d like to talk it through, book a free 15-minute chat.

Sources