What happens to what you type, what changes with a business plan, and the rule to give your team.
Resources / Is it safe to put customer data into ChatGPT?
Not in the free or personal versions. Australia's privacy regulator recommends that businesses do not enter personal information into publicly available AI tools at all,1 and OpenAI says it may use conversations in its individual plans to train its models unless you switch that off.2 Its business plans are different: by default, OpenAI does not train on business data.3 Even then, a customer's details need a good reason to be in there, and most of the time they do not.
The regulator's advice for public AI tools
Not trained on by default
New privacy policy rules for automated decisions
In the plans for individuals, which covers Free, Go, Plus and Pro, OpenAI says it may use your content to train its models. You can turn that off under Settings, Data controls, by switching off Improve the model for everyone.2 One catch: giving a response a thumbs up or down can still send that whole conversation for training, even with the setting off.
Temporary chats are not used for training and do not appear in your history, but OpenAI may keep them for up to 30 days for safety reasons.2 So a temporary chat is not a way to make customer data disappear.
The bigger issue is not training at all. Once a customer's details are in a tool your business does not control, they have left your business, and you can no longer say who can see them or how long they are kept.
OpenAI's business plans, which it currently calls ChatGPT Business and ChatGPT Enterprise, work differently. OpenAI says it does not use business data for training by default, and on Enterprise, your workspace administrator controls how long data is kept.3
Microsoft's Copilot is similar when staff sign in with a work account. Microsoft says prompts and responses under its enterprise data protection are not used to train its foundation models.4 If your business already runs on Microsoft 365, that is often the simplest option with those protections, because it sits inside the same accounts you already manage.
A business plan fixes the training problem and gives you control. It does not decide what is appropriate to put in, which is still your job.
The OAIC's guidance is that putting personal information into an AI tool is a use or a disclosure under the Privacy Act, and the ordinary rules apply: it should match the reason you collected the information, or be something the customer would reasonably expect, or have their consent.1 Pasting a customer's complaint into a chatbot to draft a reply is a very different thing from pasting in their medical history.
Many small businesses are exempt from the Privacy Act: the threshold is annual turnover of $3 million or less. But some are covered regardless of turnover, including health service providers.5 And exempt or not, customers expect the same care, and our guide to the Privacy Act changes explains why the exemption protects less than it used to.
The same question applies to AI features built into software you already pay for: email platforms, design tools, customer management systems and accounting software increasingly have an AI button. Each one sends data somewhere, so check its AI or data settings the same way you would check ChatGPT's before customer details go through it.
From 10 December 2026, businesses covered by the Privacy Act must say in their privacy policy if they use a computer program with personal information to make decisions that could significantly affect a person's rights or interests.6
Most small business use of ChatGPT is drafting and summarising, not making decisions about people, so this will not apply to much of it. It does apply if you use AI to decide, or substantially help decide, things like who is approved for credit, who is shortlisted for a job, or who is offered a service.
The OAIC also recommends that customer-facing AI, such as a chatbot on your website, is clearly identified as AI, and that your privacy policy says you use it.1 That is good practice whether or not the new rule applies to you.
Write down one rule and give it to everyone: no customer names, contact details, account numbers or anything sensitive goes into a personal or free AI tool. Health, financial and identity information does not go into any AI tool without a clear reason and approval.
Most useful work does not need the personal details anyway. Replace names with Customer A, remove addresses and numbers, and ask for the draft. The result is just as good and nothing has left the business.
Then give people the right tool, a business plan or Copilot on their work account, so they are not tempted to use a personal one. Staff who have been told no without an alternative tend to keep doing it quietly.
Do not panic, and do not make it a disciplinary matter by default; most people do this because no one told them not to. Find out what was shared and in which tool, delete the conversations, and switch off training in that account's data controls.
If what was shared was sensitive, such as health information or identity documents, get advice on whether you have any obligations to the affected customers. This article is general information, not legal advice for your situation.
Then use it as the reason to write the one-page rule above. It is the cheapest fix available, and it is the thing a customer or regulator would ask about first.
How can AI actually help my small business (and where's the catch)?: The overview: where AI helps, where it bites, and the one rule.
Does the new Privacy Act apply to my small business?: The turnover exemption, what it no longer protects, and the December 2026 changes.
Deepfakes and AI scams: how do I protect my small business?: The other side of AI: the scams it makes more convincing.
Not into the free or personal versions. The OAIC recommends businesses do not enter personal information into publicly available AI tools, and OpenAI may use individual-plan conversations for training unless you opt out. Business plans are not trained on by default, but only use customer details when there is a clear reason.
On ChatGPT Business and Enterprise, OpenAI says it does not train on business data by default. On the individual plans (Free, Go, Plus and Pro) it may, unless you turn off Improve the model for everyone in Settings, Data controls.
Yes, for businesses the Act covers. The OAIC treats putting personal information into an AI tool as a use or disclosure, so it must fit the purpose you collected it for, or the customer's reasonable expectations, or have consent. Businesses with turnover of $3 million or less are often exempt, with exceptions such as health service providers.
From 10 December 2026, businesses covered by the Privacy Act must explain in their privacy policy if they use a computer program with personal information to make decisions that could significantly affect people. Everyday drafting and summarising usually is not that kind of decision.
At minimum: no customer names, contact details or sensitive information in personal or free AI tools; use the business-approved tool; remove personal details before asking for a draft; and check anything AI produces before it goes to a customer. One page is enough.
Tell us which AI tools your team uses now. We will recommend the right business tool, write the one-page rule with you, and set it up so customer data stays in your control. Free 30-minute call.
Prefer to talk? Call (02) 9053 8789.
Belivanis Information Systems