Current at 31 October 2026. This is general information about the law, not legal advice for your business.
This is the short version for a business whose staff use ChatGPT, Copilot, Claude or Gemini at work. The Office of the Australian Information Commissioner (OAIC) is plain about the starting point: the Privacy Act applies to all uses of AI involving personal information. Its guidance on privacy and the use of commercially available AI products, published on 21 October 2024 and updated on 17 January 2025, covers paid business tools and AI products which are freely available, such as publicly accessible AI chatbots. The OAIC has separate guidance for developers who build or train models, which most businesses will not need. For the rest of what an Australian business has to do about AI, see what is actually required.
Does the Privacy Act apply to your business?
Most small businesses are not covered. A business is a small business if its annual turnover for the previous financial year is $3,000,000 or less (section 6D), and section 6C leaves small business operators out of the organisations the Act binds. The OAIC adds that annual turnover for the purposes of the Privacy Act includes all income from all sources, and the Act itself counts what is earned in the year in the course of the business (section 6DA).
The threshold has exceptions. A business is still covered if a business it carries on has had an annual turnover of more than $3,000,000 in any earlier financial year since it started, or since section 6D commenced if that is later, even when last year's turnover was lower (section 6D(4)(a)). Regardless of turnover, the Act also covers a health service provider, a business trading in personal information, a contractor providing services under a Commonwealth contract, a credit reporting body, a business related to one the Act covers and a business that has opted in, among others. Health service is read widely: the OAIC's list includes allied health professionals; complementary therapists; child care centres; private schools.
The exemption may not last, though nothing has changed yet. In its response to the Privacy Act Review Report, released on 28 September 2023, the government agreed in principle to remove it, but only after further consultation with small business. In the compilation of the Act in force on 4 June 2026, the $3,000,000 test is still there. Even for a business it does not cover, the OAIC says: "we recommend you protect any personal information you hold". Your client contracts may also ask for more than the Act does, so the rest of this is worth reading either way.
Can you put customer information into ChatGPT or Copilot?
For public tools the OAIC's advice is not to: it recommends that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools. That is advice, not a rule in the Act. If the Act covers you, these rules apply.
APP 6: use it for the reason you collected it. If personal information goes into an AI system, APP 6 requires you to only use or disclose the information for the primary purpose for which it was collected, unless the person consented, or they would reasonably expect the other use and it is related to the original purpose (directly related, for sensitive information), among other exceptions. The OAIC says a secondary use may be within an individual's reasonable expectations if it was expressly outlined in a notice at the time of collection and in your business's privacy policy. Drafting a reply to a customer's own enquiry, the example at the top, sits much closer to the original purpose than running your whole customer list through a tool to look for patterns.
APP 3: what the tool produces counts too. If an AI tool is used to generate or infer personal information, the OAIC says this is a collection of personal information and must comply with APP 3. That includes the wrong answers: inferred, incorrect or artificially generated information produced by AI models (such as hallucinations and deepfakes) about an identifiable person is personal information. The OAIC also says not to record or treat these outputs as facts.
The tool's own terms. The OAIC warns that some commercial AI products will include terms or settings that allow the product owner to collect the data input by customers for further training, so that personal information could surface in an answer to another user. The measures it suggests include turning off features that would disclose personal information and prohibiting staff from entering personal information. Which business plans train on your data by default is set out in our comparison of ChatGPT, Copilot and Claude.
Is sending customer data to an overseas AI tool a disclosure?
It turns on control. The OAIC says putting personal information into an AI system may be considered either a use of personal information (if the data stays within your organisation's control) or a disclosure if it is made accessible to others outside your organisation and has been released from your organisation's effective control. Its guidelines on APP 8 say that giving personal information to a contractor overseas is, in most circumstances, a disclosure.
A disclosure to someone outside Australia brings in APP 8. Before it happens you must take such steps as are reasonable in the circumstances to ensure that the overseas recipient does not breach the Australian Privacy Principles, unless an exception in APP 8.2 applies. If the recipient is not itself bound by the Australian Privacy Principles and does something that would breach them, section 16C treats that as a breach by your business. The OAIC does not rule on any named product. Control is set in the provider's terms: what it may do with your inputs, who else handles them, and where. Where each provider stores and processes data is set out in the storage section of our ChatGPT, Copilot and Claude comparison.
What do you have to tell customers?
APP 1 requires a clearly expressed and up-to-date policy about how you manage personal information. APP 5 requires reasonable steps to tell people the relevant matters at or before the time you collect their information, or as soon as practicable after. The OAIC's advice is that businesses should update their privacy policies and notifications with clear and transparent information about their use of AI, and that a public facing AI tool such as a chatbot should be clearly identified as one. In its worked example it notes that it may be difficult to establish reasonable expectations if customers were not specifically notified that their information may go to AI developers.
What changed in 2024, and what starts on 10 December 2026
The Privacy and Other Legislation Amendment Act 2024 received Royal Assent on 10 December 2024. Four parts of it matter to a business using AI tools.
Security steps now include named measures. From 11 December 2024, the reasonable steps APP 11 asks for to protect personal information include technical and organisational measures. Access settings on an AI tool, and a written rule about what staff may paste into it, both fit that description.
Infringement notices for a privacy policy. From the same date, failing to have a privacy policy, or one that does not contain what APP 1.4 requires, is among the breaches that are subject to an infringement notice.
A statutory tort. The Act added Schedule 2 to the Privacy Act, which lets an individual sue for a serious invasion of privacy, including misusing information that relates to the plaintiff. It commenced on 10 June 2025. The invasion must be intentional or reckless and serious. Its exemptions cover journalists, government agencies, law enforcement and intelligence bodies, and people under 18, and there is no turnover test, so it can reach a business the Australian Privacy Principles do not cover.
Automated decisions, from 10 December 2026. From 10 December 2026, if a computer program makes a decision, or does something substantially and directly related to making one, that could reasonably be expected to significantly affect the rights or interests of an individual, using their personal information, your privacy policy must say so. It must set out the kinds of personal information used and the kinds of decisions made. The Act's examples include a decision that affects the individual's rights under a contract, agreement or arrangement and one that affects access to a significant service or support. The rule applies in relation to decisions made after the commencement, even if the system was set up earlier. If you use an AI tool to screen applications or set a customer's terms, check it against this rule before 10 December 2026.
A checklist before customer details go into an AI tool
- Check whether the Act covers you against the turnover test and the exceptions. Don't assume either way.
- Use a business plan with training on your inputs switched off, and find out where the provider processes your data.
- Write down which tools are approved and what may go into each, and name the person who owns that list.
- Update your privacy policy and collection notice to say you use AI tools, and that information may go to their providers, including overseas.
- Treat anything a tool says about a person as a draft. A person checks it before anyone relies on it.
- Keep sensitive information out unless you have the person's consent and a real need. In the Act, sensitive information includes health information, and information or an opinion about a person's racial or ethnic origin, religious beliefs, sexual orientation, criminal record or trade union membership.
- If the Act covers you and a program makes decisions, or does something substantially and directly related to making them, that could significantly affect people, add them to your privacy policy before 10 December 2026.
Common questions
Does the Privacy Act apply to a sole trader?
The same test applies. The Act's definition of an organisation includes an individual, so a sole trader is covered unless they are a small business operator, which turns on the $3,000,000 turnover test and the exceptions above.
Is information an AI tool generates about a customer personal information?
Yes, if it is about someone who can be identified. The OAIC says inferred, incorrect or artificially generated information about an identified or reasonably identifiable individual is personal information, even when it is wrong.
Do you need a customer's consent to use AI on their information?
Not always. Under APP 6 you can use it for the purpose you collected it, or for a related purpose the person would reasonably expect. For sensitive information the OAIC says make sure you have the individual's consent before you collect or generate it with AI.
Do you have to report it if an AI tool exposes customer data?
If the Act covers you, you must notify the people affected and the OAIC when a data breach is likely to result in serious harm to someone whose information is involved. If you only suspect a breach, section 26WH requires you to take all reasonable steps to finish assessing it within 30 days. A breach at an overseas provider can count as yours: where APP 8.1 applied when you sent the information, section 26WC treats your business as still holding it.